Lamapi commited on
Commit
759bd70
·
verified ·
1 Parent(s): 22deb15

Publish Next-Federation model

Browse files
Files changed (8) hide show
  1. README.md +66 -0
  2. evaluation.json +1654 -0
  3. fed_core.py +164 -0
  4. inference.py +53 -0
  5. model.json +1 -0
  6. nlrt.py +226 -0
  7. preview.png +0 -0
  8. selftest.json +1 -0
README.md ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ library_name: numpy
4
+ tags:
5
+ - federated-learning
6
+ - simulation
7
+ - next-lab
8
+ ---
9
+
10
+ # Next Federation
11
+
12
+ Next Federation trains one digit classifier across **10 clients whose data are distributed differently**
13
+ (Dirichlet label skew) and who sometimes **drop out**. It compares FedAvg, FedProx and local-only training on global
14
+ accuracy, the **worst client's** accuracy on its own data, and communication traffic.
15
+ **This is a simulation: all clients run in one process; nothing was deployed to real devices.**
16
+
17
+ **Space:** https://huggingface.co/spaces/thelamapi/Next-Federation — change skew, dropout, clients and rounds; all three methods retrain in the browser.
18
+
19
+ ## Results (5 seeds per setting, 40 rounds, 2 local epochs, lr 0.3, FedProx μ 0.01)
20
+
21
+ | Setting | Method | Global test accuracy | Worst client | Mean client | Traffic (MB) |
22
+ |---|---|---:|---:|---:|---:|
23
+ | alpha=0.1, dropout=0.0 | FedAvg | 0.934 ± 0.011 | 0.784 ± 0.100 | 0.940 | 7.7 |
24
+ | alpha=0.1, dropout=0.0 | FedProx | 0.934 ± 0.011 | 0.784 ± 0.100 | 0.940 | 7.7 |
25
+ | alpha=0.1, dropout=0.0 | Local only | 0.357 ± 0.011 | 0.448 ± 0.391 | 0.920 | 0.0 |
26
+ | alpha=0.1, dropout=0.3 | FedAvg | 0.928 ± 0.019 | 0.776 ± 0.090 | 0.934 | 6.5 |
27
+ | alpha=0.1, dropout=0.3 | FedProx | 0.929 ± 0.018 | 0.769 ± 0.090 | 0.932 | 6.5 |
28
+ | alpha=0.1, dropout=0.3 | Local only | 0.357 ± 0.011 | 0.448 ± 0.391 | 0.920 | 0.0 |
29
+ | alpha=0.5, dropout=0.0 | FedAvg | 0.953 ± 0.004 | 0.866 ± 0.031 | 0.947 | 7.7 |
30
+ | alpha=0.5, dropout=0.0 | FedProx | 0.952 ± 0.003 | 0.866 ± 0.031 | 0.948 | 7.7 |
31
+ | alpha=0.5, dropout=0.0 | Local only | 0.657 ± 0.019 | 0.674 ± 0.092 | 0.881 | 0.0 |
32
+ | alpha=0.5, dropout=0.3 | FedAvg | 0.952 ± 0.005 | 0.860 ± 0.025 | 0.949 | 6.5 |
33
+ | alpha=0.5, dropout=0.3 | FedProx | 0.952 ± 0.006 | 0.860 ± 0.025 | 0.949 | 6.5 |
34
+ | alpha=0.5, dropout=0.3 | Local only | 0.657 ± 0.019 | 0.674 ± 0.092 | 0.881 | 0.0 |
35
+ | alpha=5.0, dropout=0.0 | FedAvg | 0.961 ± 0.004 | 0.880 ± 0.013 | 0.956 | 7.7 |
36
+ | alpha=5.0, dropout=0.0 | FedProx | 0.960 ± 0.005 | 0.880 ± 0.013 | 0.955 | 7.7 |
37
+ | alpha=5.0, dropout=0.0 | Local only | 0.871 ± 0.006 | 0.755 ± 0.074 | 0.883 | 0.0 |
38
+ | alpha=5.0, dropout=0.3 | FedAvg | 0.960 ± 0.005 | 0.890 ± 0.013 | 0.955 | 6.5 |
39
+ | alpha=5.0, dropout=0.3 | FedProx | 0.959 ± 0.005 | 0.890 ± 0.013 | 0.955 | 6.5 |
40
+ | alpha=5.0, dropout=0.3 | Local only | 0.871 ± 0.006 | 0.755 ± 0.074 | 0.883 | 0.0 |
41
+
42
+ Global test set: 360 digits never used for training or tuning. "Worst client" is the lowest accuracy over
43
+ clients on their own held-out 20%. Local-only global accuracy is the mean of the 10 client models.
44
+ Traffic counts float32 model downloads to every client every round plus uploads that arrive.
45
+
46
+ ![Benchmark](preview.png)
47
+
48
+ ## Method
49
+ * Model: MLP 64→32→10 with ReLU, softmax cross-entropy, minibatch SGD (batch 20).
50
+ * FedAvg: sample-weighted average of the updates that arrive. FedProx adds (μ/2)‖w − w_global‖² to every local step.
51
+ * Dropout: a client's update is lost with probability p (its download still counts).
52
+ * lr and μ were chosen on a separate validation set with development seeds 100–101 (lr scores {0.05: 0.9027777777777778, 0.1: 0.9361111111111111, 0.3: 0.9527777777777777}, μ scores {0.01: 0.9527777777777777, 0.1: 0.95, 1.0: 0.9333333333333333}).
53
+ * The Space runs a line-by-line JavaScript port; Node parity on 4 fixed runs: max abs difference 0.
54
+ * Released weights (`model.json`): FedProx global model from α=0.5, dropout 0.3, seed 0.
55
+
56
+ ## Usage
57
+ ```python
58
+ from inference import load, predict
59
+ spec = load()
60
+ probs = predict(spec, pixels) # pixels: (N, 64) ints 0..16
61
+ ```
62
+
63
+ ## Limitations
64
+ * Simulated clients on a small, clean dataset; real federations have system heterogeneity, stragglers and privacy mechanisms (none here).
65
+ * Dropout is independent per client and round; real outages are correlated.
66
+ * Recorded on NVIDIA L4. The numbers describe this simulation only.
evaluation.json ADDED
@@ -0,0 +1,1654 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "project": "Next Federation",
3
+ "profile": "full",
4
+ "seed": 73,
5
+ "device": {
6
+ "python": "3.13.16",
7
+ "platform": "Linux-6.6.159+-x86_64-with-glibc2.39",
8
+ "cpu_count": 12,
9
+ "numpy": "2.4.6",
10
+ "torch": "2.14.1+cu130",
11
+ "cuda": true,
12
+ "gpu": "NVIDIA L4",
13
+ "colab": true
14
+ },
15
+ "data": {
16
+ "name": "Optical Recognition of Handwritten Digits",
17
+ "source": "UCI id 80 via sklearn.datasets.load_digits",
18
+ "license": "CC BY 4.0",
19
+ "sklearn_version": "1.6.1",
20
+ "split": {
21
+ "global_test": 360,
22
+ "global_validation": 180,
23
+ "client_pool": 1257,
24
+ "clients": 10,
25
+ "client_partition": "Dirichlet(alpha) label skew per class; 20% of each client's data is its local test set"
26
+ }
27
+ },
28
+ "simulation": "All clients simulated in one process. No real devices, networks or secure aggregation.",
29
+ "hyper": {
30
+ "epochs": 2,
31
+ "lr": 0.3,
32
+ "batch": 20,
33
+ "mu": 0.01
34
+ },
35
+ "hyper_selection": {
36
+ "split": "global validation (180)",
37
+ "lr_scores": {
38
+ "0.05": 0.9027777777777778,
39
+ "0.1": 0.9361111111111111,
40
+ "0.3": 0.9527777777777777
41
+ },
42
+ "mu_scores": {
43
+ "0.01": 0.9527777777777777,
44
+ "0.1": 0.95,
45
+ "1.0": 0.9333333333333333
46
+ },
47
+ "seeds": [
48
+ 100,
49
+ 101
50
+ ]
51
+ },
52
+ "protocol": {
53
+ "rounds": 40,
54
+ "clients": 10,
55
+ "seeds": [
56
+ 0,
57
+ 1,
58
+ 2,
59
+ 3,
60
+ 4
61
+ ],
62
+ "alphas": [
63
+ 0.1,
64
+ 0.5,
65
+ 5.0
66
+ ],
67
+ "dropouts": [
68
+ 0.0,
69
+ 0.3
70
+ ],
71
+ "local_only": "each client trains alone for the same rounds x epochs; global accuracy = mean of client models on the global test set"
72
+ },
73
+ "results": {
74
+ "alpha=0.1, dropout=0.0": {
75
+ "fedavg": {
76
+ "global_acc": {
77
+ "mean": 0.9344444444444445,
78
+ "std": 0.011194134266449936
79
+ },
80
+ "worst_client": {
81
+ "mean": 0.7838455013455013,
82
+ "std": 0.10041913337230032
83
+ },
84
+ "mean_client": {
85
+ "mean": 0.9403127647410752,
86
+ "std": 0.022132639952130453
87
+ },
88
+ "comm_MB": {
89
+ "mean": 7.712000000000001,
90
+ "std": 8.881784197001252e-16
91
+ },
92
+ "seeds": 5
93
+ },
94
+ "fedprox": {
95
+ "global_acc": {
96
+ "mean": 0.933888888888889,
97
+ "std": 0.0111665284679121
98
+ },
99
+ "worst_client": {
100
+ "mean": 0.7838455013455013,
101
+ "std": 0.10041913337230032
102
+ },
103
+ "mean_client": {
104
+ "mean": 0.9403127647410752,
105
+ "std": 0.022132639952130453
106
+ },
107
+ "comm_MB": {
108
+ "mean": 7.712000000000001,
109
+ "std": 8.881784197001252e-16
110
+ },
111
+ "seeds": 5
112
+ },
113
+ "local": {
114
+ "global_acc": {
115
+ "mean": 0.35716666666666663,
116
+ "std": 0.010702774533552008
117
+ },
118
+ "worst_client": {
119
+ "mean": 0.44848484848484854,
120
+ "std": 0.39120261472382634
121
+ },
122
+ "mean_client": {
123
+ "mean": 0.9199442628749219,
124
+ "std": 0.04357931105513403
125
+ },
126
+ "comm_MB": {
127
+ "mean": 0.0,
128
+ "std": 0.0
129
+ },
130
+ "seeds": 5
131
+ }
132
+ },
133
+ "alpha=0.1, dropout=0.3": {
134
+ "fedavg": {
135
+ "global_acc": {
136
+ "mean": 0.9283333333333333,
137
+ "std": 0.019196836159078856
138
+ },
139
+ "worst_client": {
140
+ "mean": 0.776055501102671,
141
+ "std": 0.09049021965912739
142
+ },
143
+ "mean_client": {
144
+ "mean": 0.934044279466,
145
+ "std": 0.020745206486618272
146
+ },
147
+ "comm_MB": {
148
+ "mean": 6.530136000000001,
149
+ "std": 0.04627199999999989
150
+ },
151
+ "seeds": 5
152
+ },
153
+ "fedprox": {
154
+ "global_acc": {
155
+ "mean": 0.9294444444444444,
156
+ "std": 0.01813869746368276
157
+ },
158
+ "worst_client": {
159
+ "mean": 0.7689162864162864,
160
+ "std": 0.09032289462840334
161
+ },
162
+ "mean_client": {
163
+ "mean": 0.9318520872738076,
164
+ "std": 0.020717394394941257
165
+ },
166
+ "comm_MB": {
167
+ "mean": 6.530136000000001,
168
+ "std": 0.04627199999999989
169
+ },
170
+ "seeds": 5
171
+ },
172
+ "local": {
173
+ "global_acc": {
174
+ "mean": 0.35716666666666663,
175
+ "std": 0.010702774533552008
176
+ },
177
+ "worst_client": {
178
+ "mean": 0.44848484848484854,
179
+ "std": 0.39120261472382634
180
+ },
181
+ "mean_client": {
182
+ "mean": 0.9199442628749219,
183
+ "std": 0.04357931105513403
184
+ },
185
+ "comm_MB": {
186
+ "mean": 0.0,
187
+ "std": 0.0
188
+ },
189
+ "seeds": 5
190
+ }
191
+ },
192
+ "alpha=0.5, dropout=0.0": {
193
+ "fedavg": {
194
+ "global_acc": {
195
+ "mean": 0.9533333333333334,
196
+ "std": 0.004082482904638658
197
+ },
198
+ "worst_client": {
199
+ "mean": 0.8656385281385282,
200
+ "std": 0.030617187247437064
201
+ },
202
+ "mean_client": {
203
+ "mean": 0.9473659555003355,
204
+ "std": 0.009621856486895888
205
+ },
206
+ "comm_MB": {
207
+ "mean": 7.712000000000001,
208
+ "std": 8.881784197001252e-16
209
+ },
210
+ "seeds": 5
211
+ },
212
+ "fedprox": {
213
+ "global_acc": {
214
+ "mean": 0.9522222222222222,
215
+ "std": 0.0032394177193585307
216
+ },
217
+ "worst_client": {
218
+ "mean": 0.8656385281385282,
219
+ "std": 0.030617187247437064
220
+ },
221
+ "mean_client": {
222
+ "mean": 0.9482355207177268,
223
+ "std": 0.00997630125698648
224
+ },
225
+ "comm_MB": {
226
+ "mean": 7.712000000000001,
227
+ "std": 8.881784197001252e-16
228
+ },
229
+ "seeds": 5
230
+ },
231
+ "local": {
232
+ "global_acc": {
233
+ "mean": 0.6569444444444444,
234
+ "std": 0.018866817169927146
235
+ },
236
+ "worst_client": {
237
+ "mean": 0.6737698412698412,
238
+ "std": 0.09179081814534823
239
+ },
240
+ "mean_client": {
241
+ "mean": 0.8811313738580184,
242
+ "std": 0.015293929817571052
243
+ },
244
+ "comm_MB": {
245
+ "mean": 0.0,
246
+ "std": 0.0
247
+ },
248
+ "seeds": 5
249
+ }
250
+ },
251
+ "alpha=0.5, dropout=0.3": {
252
+ "fedavg": {
253
+ "global_acc": {
254
+ "mean": 0.9516666666666668,
255
+ "std": 0.004513354669242236
256
+ },
257
+ "worst_client": {
258
+ "mean": 0.8596861471861473,
259
+ "std": 0.024766530385624597
260
+ },
261
+ "mean_client": {
262
+ "mean": 0.949013181562629,
263
+ "std": 0.00932022337777485
264
+ },
265
+ "comm_MB": {
266
+ "mean": 6.530136000000001,
267
+ "std": 0.04627199999999989
268
+ },
269
+ "seeds": 5
270
+ },
271
+ "fedprox": {
272
+ "global_acc": {
273
+ "mean": 0.9516666666666668,
274
+ "std": 0.0057197945227705875
275
+ },
276
+ "worst_client": {
277
+ "mean": 0.8596861471861473,
278
+ "std": 0.024766530385624597
279
+ },
280
+ "mean_client": {
281
+ "mean": 0.949013181562629,
282
+ "std": 0.00932022337777485
283
+ },
284
+ "comm_MB": {
285
+ "mean": 6.530136000000001,
286
+ "std": 0.04627199999999989
287
+ },
288
+ "seeds": 5
289
+ },
290
+ "local": {
291
+ "global_acc": {
292
+ "mean": 0.6569444444444444,
293
+ "std": 0.018866817169927146
294
+ },
295
+ "worst_client": {
296
+ "mean": 0.6737698412698412,
297
+ "std": 0.09179081814534823
298
+ },
299
+ "mean_client": {
300
+ "mean": 0.8811313738580184,
301
+ "std": 0.015293929817571052
302
+ },
303
+ "comm_MB": {
304
+ "mean": 0.0,
305
+ "std": 0.0
306
+ },
307
+ "seeds": 5
308
+ }
309
+ },
310
+ "alpha=5.0, dropout=0.0": {
311
+ "fedavg": {
312
+ "global_acc": {
313
+ "mean": 0.961111111111111,
314
+ "std": 0.003928371006591917
315
+ },
316
+ "worst_client": {
317
+ "mean": 0.8803085299455535,
318
+ "std": 0.01269169078142293
319
+ },
320
+ "mean_client": {
321
+ "mean": 0.9561164867853661,
322
+ "std": 0.010027910118678741
323
+ },
324
+ "comm_MB": {
325
+ "mean": 7.712000000000001,
326
+ "std": 8.881784197001252e-16
327
+ },
328
+ "seeds": 5
329
+ },
330
+ "fedprox": {
331
+ "global_acc": {
332
+ "mean": 0.96,
333
+ "std": 0.00451335466924222
334
+ },
335
+ "worst_client": {
336
+ "mean": 0.8803085299455535,
337
+ "std": 0.01269169078142293
338
+ },
339
+ "mean_client": {
340
+ "mean": 0.955316486785366,
341
+ "std": 0.01068387952282781
342
+ },
343
+ "comm_MB": {
344
+ "mean": 7.712000000000001,
345
+ "std": 8.881784197001252e-16
346
+ },
347
+ "seeds": 5
348
+ },
349
+ "local": {
350
+ "global_acc": {
351
+ "mean": 0.8706666666666667,
352
+ "std": 0.005606984180906709
353
+ },
354
+ "worst_client": {
355
+ "mean": 0.7554372041214147,
356
+ "std": 0.07390787762854817
357
+ },
358
+ "mean_client": {
359
+ "mean": 0.8833721562105143,
360
+ "std": 0.014421833115630082
361
+ },
362
+ "comm_MB": {
363
+ "mean": 0.0,
364
+ "std": 0.0
365
+ },
366
+ "seeds": 5
367
+ }
368
+ },
369
+ "alpha=5.0, dropout=0.3": {
370
+ "fedavg": {
371
+ "global_acc": {
372
+ "mean": 0.96,
373
+ "std": 0.0051520102752754164
374
+ },
375
+ "worst_client": {
376
+ "mean": 0.8900758950668207,
377
+ "std": 0.01326496718368544
378
+ },
379
+ "mean_client": {
380
+ "mean": 0.9552134507435783,
381
+ "std": 0.009768884623512993
382
+ },
383
+ "comm_MB": {
384
+ "mean": 6.530136000000001,
385
+ "std": 0.04627199999999989
386
+ },
387
+ "seeds": 5
388
+ },
389
+ "fedprox": {
390
+ "global_acc": {
391
+ "mean": 0.9594444444444445,
392
+ "std": 0.004843221048378552
393
+ },
394
+ "worst_client": {
395
+ "mean": 0.8900758950668207,
396
+ "std": 0.01326496718368544
397
+ },
398
+ "mean_client": {
399
+ "mean": 0.9552134507435783,
400
+ "std": 0.009768884623512993
401
+ },
402
+ "comm_MB": {
403
+ "mean": 6.530136000000001,
404
+ "std": 0.04627199999999989
405
+ },
406
+ "seeds": 5
407
+ },
408
+ "local": {
409
+ "global_acc": {
410
+ "mean": 0.8706666666666667,
411
+ "std": 0.005606984180906709
412
+ },
413
+ "worst_client": {
414
+ "mean": 0.7554372041214147,
415
+ "std": 0.07390787762854817
416
+ },
417
+ "mean_client": {
418
+ "mean": 0.8833721562105143,
419
+ "std": 0.014421833115630082
420
+ },
421
+ "comm_MB": {
422
+ "mean": 0.0,
423
+ "std": 0.0
424
+ },
425
+ "seeds": 5
426
+ }
427
+ }
428
+ },
429
+ "lowest_worst_client_runs": [
430
+ {
431
+ "alpha": 0.1,
432
+ "p_drop": 0.0,
433
+ "seed": 2,
434
+ "method": "fedavg",
435
+ "worst_client": 0.6,
436
+ "client_sizes": [
437
+ 5,
438
+ 37,
439
+ 140,
440
+ 264,
441
+ 258,
442
+ 132,
443
+ 36,
444
+ 21,
445
+ 58,
446
+ 57
447
+ ]
448
+ },
449
+ {
450
+ "alpha": 0.1,
451
+ "p_drop": 0.0,
452
+ "seed": 2,
453
+ "method": "fedprox",
454
+ "worst_client": 0.6,
455
+ "client_sizes": [
456
+ 5,
457
+ 37,
458
+ 140,
459
+ 264,
460
+ 258,
461
+ 132,
462
+ 36,
463
+ 21,
464
+ 58,
465
+ 57
466
+ ]
467
+ },
468
+ {
469
+ "alpha": 0.1,
470
+ "p_drop": 0.3,
471
+ "seed": 2,
472
+ "method": "fedavg",
473
+ "worst_client": 0.6,
474
+ "client_sizes": [
475
+ 5,
476
+ 37,
477
+ 140,
478
+ 264,
479
+ 258,
480
+ 132,
481
+ 36,
482
+ 21,
483
+ 58,
484
+ 57
485
+ ]
486
+ },
487
+ {
488
+ "alpha": 0.1,
489
+ "p_drop": 0.3,
490
+ "seed": 2,
491
+ "method": "fedprox",
492
+ "worst_client": 0.6,
493
+ "client_sizes": [
494
+ 5,
495
+ 37,
496
+ 140,
497
+ 264,
498
+ 258,
499
+ 132,
500
+ 36,
501
+ 21,
502
+ 58,
503
+ 57
504
+ ]
505
+ },
506
+ {
507
+ "alpha": 0.1,
508
+ "p_drop": 0.3,
509
+ "seed": 0,
510
+ "method": "fedprox",
511
+ "worst_client": 0.7567567567567568,
512
+ "client_sizes": [
513
+ 5,
514
+ 150,
515
+ 104,
516
+ 42,
517
+ 16,
518
+ 216,
519
+ 58,
520
+ 176,
521
+ 213,
522
+ 28
523
+ ]
524
+ },
525
+ {
526
+ "alpha": 0.1,
527
+ "p_drop": 0.0,
528
+ "seed": 0,
529
+ "method": "fedavg",
530
+ "worst_client": 0.7837837837837838,
531
+ "client_sizes": [
532
+ 5,
533
+ 150,
534
+ 104,
535
+ 42,
536
+ 16,
537
+ 216,
538
+ 58,
539
+ 176,
540
+ 213,
541
+ 28
542
+ ]
543
+ },
544
+ {
545
+ "alpha": 0.1,
546
+ "p_drop": 0.0,
547
+ "seed": 0,
548
+ "method": "fedprox",
549
+ "worst_client": 0.7837837837837838,
550
+ "client_sizes": [
551
+ 5,
552
+ 150,
553
+ 104,
554
+ 42,
555
+ 16,
556
+ 216,
557
+ 58,
558
+ 176,
559
+ 213,
560
+ 28
561
+ ]
562
+ },
563
+ {
564
+ "alpha": 0.1,
565
+ "p_drop": 0.3,
566
+ "seed": 0,
567
+ "method": "fedavg",
568
+ "worst_client": 0.7924528301886793,
569
+ "client_sizes": [
570
+ 5,
571
+ 150,
572
+ 104,
573
+ 42,
574
+ 16,
575
+ 216,
576
+ 58,
577
+ 176,
578
+ 213,
579
+ 28
580
+ ]
581
+ },
582
+ {
583
+ "alpha": 0.1,
584
+ "p_drop": 0.0,
585
+ "seed": 4,
586
+ "method": "fedavg",
587
+ "worst_client": 0.8125,
588
+ "client_sizes": [
589
+ 4,
590
+ 90,
591
+ 64,
592
+ 96,
593
+ 28,
594
+ 280,
595
+ 155,
596
+ 144,
597
+ 63,
598
+ 85
599
+ ]
600
+ },
601
+ {
602
+ "alpha": 0.1,
603
+ "p_drop": 0.0,
604
+ "seed": 4,
605
+ "method": "fedprox",
606
+ "worst_client": 0.8125,
607
+ "client_sizes": [
608
+ 4,
609
+ 90,
610
+ 64,
611
+ 96,
612
+ 28,
613
+ 280,
614
+ 155,
615
+ 144,
616
+ 63,
617
+ 85
618
+ ]
619
+ }
620
+ ],
621
+ "runs": [
622
+ {
623
+ "alpha": 0.1,
624
+ "p_drop": 0.0,
625
+ "seed": 0,
626
+ "method": "fedavg",
627
+ "global_acc": 0.9138888888888889,
628
+ "worst_client": 0.7837837837837838,
629
+ "mean_client": 0.9556643000039227,
630
+ "comm_MB": 7.712,
631
+ "uploads": 400
632
+ },
633
+ {
634
+ "alpha": 0.1,
635
+ "p_drop": 0.0,
636
+ "seed": 0,
637
+ "method": "fedprox",
638
+ "global_acc": 0.9138888888888889,
639
+ "worst_client": 0.7837837837837838,
640
+ "mean_client": 0.9556643000039227,
641
+ "comm_MB": 7.712,
642
+ "uploads": 400
643
+ },
644
+ {
645
+ "alpha": 0.1,
646
+ "p_drop": 0.0,
647
+ "seed": 0,
648
+ "method": "local",
649
+ "global_acc": 0.36138888888888887,
650
+ "worst_client": 0.0,
651
+ "mean_client": 0.8896775585454831,
652
+ "comm_MB": 0.0,
653
+ "uploads": 0
654
+ },
655
+ {
656
+ "alpha": 0.1,
657
+ "p_drop": 0.0,
658
+ "seed": 1,
659
+ "method": "fedavg",
660
+ "global_acc": 0.9444444444444444,
661
+ "worst_client": 0.8181818181818182,
662
+ "mean_client": 0.9308770053475935,
663
+ "comm_MB": 7.712,
664
+ "uploads": 400
665
+ },
666
+ {
667
+ "alpha": 0.1,
668
+ "p_drop": 0.0,
669
+ "seed": 1,
670
+ "method": "fedprox",
671
+ "global_acc": 0.9444444444444444,
672
+ "worst_client": 0.8181818181818182,
673
+ "mean_client": 0.9308770053475935,
674
+ "comm_MB": 7.712,
675
+ "uploads": 400
676
+ },
677
+ {
678
+ "alpha": 0.1,
679
+ "p_drop": 0.0,
680
+ "seed": 1,
681
+ "method": "local",
682
+ "global_acc": 0.3666666666666667,
683
+ "worst_client": 0.9090909090909091,
684
+ "mean_client": 0.9781345811051694,
685
+ "comm_MB": 0.0,
686
+ "uploads": 0
687
+ },
688
+ {
689
+ "alpha": 0.1,
690
+ "p_drop": 0.0,
691
+ "seed": 2,
692
+ "method": "fedavg",
693
+ "global_acc": 0.9333333333333333,
694
+ "worst_client": 0.6,
695
+ "mean_client": 0.9271865634365634,
696
+ "comm_MB": 7.712,
697
+ "uploads": 400
698
+ },
699
+ {
700
+ "alpha": 0.1,
701
+ "p_drop": 0.0,
702
+ "seed": 2,
703
+ "method": "fedprox",
704
+ "global_acc": 0.9333333333333333,
705
+ "worst_client": 0.6,
706
+ "mean_client": 0.9271865634365634,
707
+ "comm_MB": 7.712,
708
+ "uploads": 400
709
+ },
710
+ {
711
+ "alpha": 0.1,
712
+ "p_drop": 0.0,
713
+ "seed": 2,
714
+ "method": "local",
715
+ "global_acc": 0.3683333333333333,
716
+ "worst_client": 0.0,
717
+ "mean_client": 0.8552836399711399,
718
+ "comm_MB": 0.0,
719
+ "uploads": 0
720
+ },
721
+ {
722
+ "alpha": 0.1,
723
+ "p_drop": 0.0,
724
+ "seed": 3,
725
+ "method": "fedavg",
726
+ "global_acc": 0.9444444444444444,
727
+ "worst_client": 0.9047619047619048,
728
+ "mean_client": 0.9749524203232347,
729
+ "comm_MB": 7.712,
730
+ "uploads": 400
731
+ },
732
+ {
733
+ "alpha": 0.1,
734
+ "p_drop": 0.0,
735
+ "seed": 3,
736
+ "method": "fedprox",
737
+ "global_acc": 0.9444444444444444,
738
+ "worst_client": 0.9047619047619048,
739
+ "mean_client": 0.9749524203232347,
740
+ "comm_MB": 7.712,
741
+ "uploads": 400
742
+ },
743
+ {
744
+ "alpha": 0.1,
745
+ "p_drop": 0.0,
746
+ "seed": 3,
747
+ "method": "local",
748
+ "global_acc": 0.3486111111111111,
749
+ "worst_client": 0.5,
750
+ "mean_client": 0.9253167071281995,
751
+ "comm_MB": 0.0,
752
+ "uploads": 0
753
+ },
754
+ {
755
+ "alpha": 0.1,
756
+ "p_drop": 0.0,
757
+ "seed": 4,
758
+ "method": "fedavg",
759
+ "global_acc": 0.9361111111111111,
760
+ "worst_client": 0.8125,
761
+ "mean_client": 0.9128835345940609,
762
+ "comm_MB": 7.712,
763
+ "uploads": 400
764
+ },
765
+ {
766
+ "alpha": 0.1,
767
+ "p_drop": 0.0,
768
+ "seed": 4,
769
+ "method": "fedprox",
770
+ "global_acc": 0.9333333333333333,
771
+ "worst_client": 0.8125,
772
+ "mean_client": 0.9128835345940609,
773
+ "comm_MB": 7.712,
774
+ "uploads": 400
775
+ },
776
+ {
777
+ "alpha": 0.1,
778
+ "p_drop": 0.0,
779
+ "seed": 4,
780
+ "method": "local",
781
+ "global_acc": 0.3408333333333333,
782
+ "worst_client": 0.8333333333333334,
783
+ "mean_client": 0.9513088276246171,
784
+ "comm_MB": 0.0,
785
+ "uploads": 0
786
+ },
787
+ {
788
+ "alpha": 0.1,
789
+ "p_drop": 0.3,
790
+ "seed": 0,
791
+ "method": "fedavg",
792
+ "global_acc": 0.8944444444444445,
793
+ "worst_client": 0.7924528301886793,
794
+ "mean_client": 0.9393373531109381,
795
+ "comm_MB": 6.6034,
796
+ "uploads": 285
797
+ },
798
+ {
799
+ "alpha": 0.1,
800
+ "p_drop": 0.3,
801
+ "seed": 0,
802
+ "method": "fedprox",
803
+ "global_acc": 0.8972222222222223,
804
+ "worst_client": 0.7567567567567568,
805
+ "mean_client": 0.9339319477055327,
806
+ "comm_MB": 6.6034,
807
+ "uploads": 285
808
+ },
809
+ {
810
+ "alpha": 0.1,
811
+ "p_drop": 0.3,
812
+ "seed": 0,
813
+ "method": "local",
814
+ "global_acc": 0.36138888888888887,
815
+ "worst_client": 0.0,
816
+ "mean_client": 0.8896775585454831,
817
+ "comm_MB": 0.0,
818
+ "uploads": 0
819
+ },
820
+ {
821
+ "alpha": 0.1,
822
+ "p_drop": 0.3,
823
+ "seed": 1,
824
+ "method": "fedavg",
825
+ "global_acc": 0.9333333333333333,
826
+ "worst_client": 0.8181818181818182,
827
+ "mean_client": 0.9326580510992276,
828
+ "comm_MB": 6.51664,
829
+ "uploads": 276
830
+ },
831
+ {
832
+ "alpha": 0.1,
833
+ "p_drop": 0.3,
834
+ "seed": 1,
835
+ "method": "fedprox",
836
+ "global_acc": 0.9333333333333333,
837
+ "worst_client": 0.8181818181818182,
838
+ "mean_client": 0.9271024955436721,
839
+ "comm_MB": 6.51664,
840
+ "uploads": 276
841
+ },
842
+ {
843
+ "alpha": 0.1,
844
+ "p_drop": 0.3,
845
+ "seed": 1,
846
+ "method": "local",
847
+ "global_acc": 0.3666666666666667,
848
+ "worst_client": 0.9090909090909091,
849
+ "mean_client": 0.9781345811051694,
850
+ "comm_MB": 0.0,
851
+ "uploads": 0
852
+ },
853
+ {
854
+ "alpha": 0.1,
855
+ "p_drop": 0.3,
856
+ "seed": 2,
857
+ "method": "fedavg",
858
+ "global_acc": 0.9361111111111111,
859
+ "worst_client": 0.6,
860
+ "mean_client": 0.9114482392607393,
861
+ "comm_MB": 6.53592,
862
+ "uploads": 278
863
+ },
864
+ {
865
+ "alpha": 0.1,
866
+ "p_drop": 0.3,
867
+ "seed": 2,
868
+ "method": "fedprox",
869
+ "global_acc": 0.9361111111111111,
870
+ "worst_client": 0.6,
871
+ "mean_client": 0.9114482392607393,
872
+ "comm_MB": 6.53592,
873
+ "uploads": 278
874
+ },
875
+ {
876
+ "alpha": 0.1,
877
+ "p_drop": 0.3,
878
+ "seed": 2,
879
+ "method": "local",
880
+ "global_acc": 0.3683333333333333,
881
+ "worst_client": 0.0,
882
+ "mean_client": 0.8552836399711399,
883
+ "comm_MB": 0.0,
884
+ "uploads": 0
885
+ },
886
+ {
887
+ "alpha": 0.1,
888
+ "p_drop": 0.3,
889
+ "seed": 3,
890
+ "method": "fedavg",
891
+ "global_acc": 0.925,
892
+ "worst_client": 0.8571428571428571,
893
+ "mean_client": 0.97019051556133,
894
+ "comm_MB": 6.53592,
895
+ "uploads": 278
896
+ },
897
+ {
898
+ "alpha": 0.1,
899
+ "p_drop": 0.3,
900
+ "seed": 3,
901
+ "method": "fedprox",
902
+ "global_acc": 0.9277777777777778,
903
+ "worst_client": 0.8571428571428571,
904
+ "mean_client": 0.97019051556133,
905
+ "comm_MB": 6.53592,
906
+ "uploads": 278
907
+ },
908
+ {
909
+ "alpha": 0.1,
910
+ "p_drop": 0.3,
911
+ "seed": 3,
912
+ "method": "local",
913
+ "global_acc": 0.3486111111111111,
914
+ "worst_client": 0.5,
915
+ "mean_client": 0.9253167071281995,
916
+ "comm_MB": 0.0,
917
+ "uploads": 0
918
+ },
919
+ {
920
+ "alpha": 0.1,
921
+ "p_drop": 0.3,
922
+ "seed": 4,
923
+ "method": "fedavg",
924
+ "global_acc": 0.9527777777777777,
925
+ "worst_client": 0.8125,
926
+ "mean_client": 0.9165872382977646,
927
+ "comm_MB": 6.4588,
928
+ "uploads": 270
929
+ },
930
+ {
931
+ "alpha": 0.1,
932
+ "p_drop": 0.3,
933
+ "seed": 4,
934
+ "method": "fedprox",
935
+ "global_acc": 0.9527777777777777,
936
+ "worst_client": 0.8125,
937
+ "mean_client": 0.9165872382977646,
938
+ "comm_MB": 6.4588,
939
+ "uploads": 270
940
+ },
941
+ {
942
+ "alpha": 0.1,
943
+ "p_drop": 0.3,
944
+ "seed": 4,
945
+ "method": "local",
946
+ "global_acc": 0.3408333333333333,
947
+ "worst_client": 0.8333333333333334,
948
+ "mean_client": 0.9513088276246171,
949
+ "comm_MB": 0.0,
950
+ "uploads": 0
951
+ },
952
+ {
953
+ "alpha": 0.5,
954
+ "p_drop": 0.0,
955
+ "seed": 0,
956
+ "method": "fedavg",
957
+ "global_acc": 0.95,
958
+ "worst_client": 0.8125,
959
+ "mean_client": 0.9525315143302949,
960
+ "comm_MB": 7.712,
961
+ "uploads": 400
962
+ },
963
+ {
964
+ "alpha": 0.5,
965
+ "p_drop": 0.0,
966
+ "seed": 0,
967
+ "method": "fedprox",
968
+ "global_acc": 0.95,
969
+ "worst_client": 0.8125,
970
+ "mean_client": 0.9525315143302949,
971
+ "comm_MB": 7.712,
972
+ "uploads": 400
973
+ },
974
+ {
975
+ "alpha": 0.5,
976
+ "p_drop": 0.0,
977
+ "seed": 0,
978
+ "method": "local",
979
+ "global_acc": 0.6613888888888889,
980
+ "worst_client": 0.5625,
981
+ "mean_client": 0.8540916569879984,
982
+ "comm_MB": 0.0,
983
+ "uploads": 0
984
+ },
985
+ {
986
+ "alpha": 0.5,
987
+ "p_drop": 0.0,
988
+ "seed": 1,
989
+ "method": "fedavg",
990
+ "global_acc": 0.9527777777777777,
991
+ "worst_client": 0.8571428571428571,
992
+ "mean_client": 0.9335846909081343,
993
+ "comm_MB": 7.712,
994
+ "uploads": 400
995
+ },
996
+ {
997
+ "alpha": 0.5,
998
+ "p_drop": 0.0,
999
+ "seed": 1,
1000
+ "method": "fedprox",
1001
+ "global_acc": 0.9527777777777777,
1002
+ "worst_client": 0.8571428571428571,
1003
+ "mean_client": 0.9335846909081343,
1004
+ "comm_MB": 7.712,
1005
+ "uploads": 400
1006
+ },
1007
+ {
1008
+ "alpha": 0.5,
1009
+ "p_drop": 0.0,
1010
+ "seed": 1,
1011
+ "method": "local",
1012
+ "global_acc": 0.6661111111111111,
1013
+ "worst_client": 0.6,
1014
+ "mean_client": 0.8874613907713828,
1015
+ "comm_MB": 0.0,
1016
+ "uploads": 0
1017
+ },
1018
+ {
1019
+ "alpha": 0.5,
1020
+ "p_drop": 0.0,
1021
+ "seed": 2,
1022
+ "method": "fedavg",
1023
+ "global_acc": 0.9527777777777777,
1024
+ "worst_client": 0.875,
1025
+ "mean_client": 0.9496210421620974,
1026
+ "comm_MB": 7.712,
1027
+ "uploads": 400
1028
+ },
1029
+ {
1030
+ "alpha": 0.5,
1031
+ "p_drop": 0.0,
1032
+ "seed": 2,
1033
+ "method": "fedprox",
1034
+ "global_acc": 0.95,
1035
+ "worst_client": 0.875,
1036
+ "mean_client": 0.9539688682490539,
1037
+ "comm_MB": 7.712,
1038
+ "uploads": 400
1039
+ },
1040
+ {
1041
+ "alpha": 0.5,
1042
+ "p_drop": 0.0,
1043
+ "seed": 2,
1044
+ "method": "local",
1045
+ "global_acc": 0.65,
1046
+ "worst_client": 0.7777777777777778,
1047
+ "mean_client": 0.8759485515420223,
1048
+ "comm_MB": 0.0,
1049
+ "uploads": 0
1050
+ },
1051
+ {
1052
+ "alpha": 0.5,
1053
+ "p_drop": 0.0,
1054
+ "seed": 3,
1055
+ "method": "fedavg",
1056
+ "global_acc": 0.9611111111111111,
1057
+ "worst_client": 0.8787878787878788,
1058
+ "mean_client": 0.9400348377934584,
1059
+ "comm_MB": 7.712,
1060
+ "uploads": 400
1061
+ },
1062
+ {
1063
+ "alpha": 0.5,
1064
+ "p_drop": 0.0,
1065
+ "seed": 3,
1066
+ "method": "fedprox",
1067
+ "global_acc": 0.9583333333333334,
1068
+ "worst_client": 0.8787878787878788,
1069
+ "mean_client": 0.9400348377934584,
1070
+ "comm_MB": 7.712,
1071
+ "uploads": 400
1072
+ },
1073
+ {
1074
+ "alpha": 0.5,
1075
+ "p_drop": 0.0,
1076
+ "seed": 3,
1077
+ "method": "local",
1078
+ "global_acc": 0.6252777777777777,
1079
+ "worst_client": 0.6428571428571429,
1080
+ "mean_client": 0.8898212843385258,
1081
+ "comm_MB": 0.0,
1082
+ "uploads": 0
1083
+ },
1084
+ {
1085
+ "alpha": 0.5,
1086
+ "p_drop": 0.0,
1087
+ "seed": 4,
1088
+ "method": "fedavg",
1089
+ "global_acc": 0.95,
1090
+ "worst_client": 0.9047619047619048,
1091
+ "mean_client": 0.9610576923076923,
1092
+ "comm_MB": 7.712,
1093
+ "uploads": 400
1094
+ },
1095
+ {
1096
+ "alpha": 0.5,
1097
+ "p_drop": 0.0,
1098
+ "seed": 4,
1099
+ "method": "fedprox",
1100
+ "global_acc": 0.95,
1101
+ "worst_client": 0.9047619047619048,
1102
+ "mean_client": 0.9610576923076923,
1103
+ "comm_MB": 7.712,
1104
+ "uploads": 400
1105
+ },
1106
+ {
1107
+ "alpha": 0.5,
1108
+ "p_drop": 0.0,
1109
+ "seed": 4,
1110
+ "method": "local",
1111
+ "global_acc": 0.6819444444444445,
1112
+ "worst_client": 0.7857142857142857,
1113
+ "mean_client": 0.8983339856501621,
1114
+ "comm_MB": 0.0,
1115
+ "uploads": 0
1116
+ },
1117
+ {
1118
+ "alpha": 0.5,
1119
+ "p_drop": 0.3,
1120
+ "seed": 0,
1121
+ "method": "fedavg",
1122
+ "global_acc": 0.9472222222222222,
1123
+ "worst_client": 0.8125,
1124
+ "mean_client": 0.9495012112999918,
1125
+ "comm_MB": 6.6034,
1126
+ "uploads": 285
1127
+ },
1128
+ {
1129
+ "alpha": 0.5,
1130
+ "p_drop": 0.3,
1131
+ "seed": 0,
1132
+ "method": "fedprox",
1133
+ "global_acc": 0.9444444444444444,
1134
+ "worst_client": 0.8125,
1135
+ "mean_client": 0.9495012112999918,
1136
+ "comm_MB": 6.6034,
1137
+ "uploads": 285
1138
+ },
1139
+ {
1140
+ "alpha": 0.5,
1141
+ "p_drop": 0.3,
1142
+ "seed": 0,
1143
+ "method": "local",
1144
+ "global_acc": 0.6613888888888889,
1145
+ "worst_client": 0.5625,
1146
+ "mean_client": 0.8540916569879984,
1147
+ "comm_MB": 0.0,
1148
+ "uploads": 0
1149
+ },
1150
+ {
1151
+ "alpha": 0.5,
1152
+ "p_drop": 0.3,
1153
+ "seed": 1,
1154
+ "method": "fedavg",
1155
+ "global_acc": 0.9472222222222222,
1156
+ "worst_client": 0.8571428571428571,
1157
+ "mean_client": 0.93468186959152,
1158
+ "comm_MB": 6.51664,
1159
+ "uploads": 276
1160
+ },
1161
+ {
1162
+ "alpha": 0.5,
1163
+ "p_drop": 0.3,
1164
+ "seed": 1,
1165
+ "method": "fedprox",
1166
+ "global_acc": 0.9472222222222222,
1167
+ "worst_client": 0.8571428571428571,
1168
+ "mean_client": 0.93468186959152,
1169
+ "comm_MB": 6.51664,
1170
+ "uploads": 276
1171
+ },
1172
+ {
1173
+ "alpha": 0.5,
1174
+ "p_drop": 0.3,
1175
+ "seed": 1,
1176
+ "method": "local",
1177
+ "global_acc": 0.6661111111111111,
1178
+ "worst_client": 0.6,
1179
+ "mean_client": 0.8874613907713828,
1180
+ "comm_MB": 0.0,
1181
+ "uploads": 0
1182
+ },
1183
+ {
1184
+ "alpha": 0.5,
1185
+ "p_drop": 0.3,
1186
+ "seed": 2,
1187
+ "method": "fedavg",
1188
+ "global_acc": 0.9555555555555556,
1189
+ "worst_client": 0.875,
1190
+ "mean_client": 0.9581355349157207,
1191
+ "comm_MB": 6.53592,
1192
+ "uploads": 278
1193
+ },
1194
+ {
1195
+ "alpha": 0.5,
1196
+ "p_drop": 0.3,
1197
+ "seed": 2,
1198
+ "method": "fedprox",
1199
+ "global_acc": 0.9583333333333334,
1200
+ "worst_client": 0.875,
1201
+ "mean_client": 0.9581355349157207,
1202
+ "comm_MB": 6.53592,
1203
+ "uploads": 278
1204
+ },
1205
+ {
1206
+ "alpha": 0.5,
1207
+ "p_drop": 0.3,
1208
+ "seed": 2,
1209
+ "method": "local",
1210
+ "global_acc": 0.65,
1211
+ "worst_client": 0.7777777777777778,
1212
+ "mean_client": 0.8759485515420223,
1213
+ "comm_MB": 0.0,
1214
+ "uploads": 0
1215
+ },
1216
+ {
1217
+ "alpha": 0.5,
1218
+ "p_drop": 0.3,
1219
+ "seed": 3,
1220
+ "method": "fedavg",
1221
+ "global_acc": 0.9583333333333334,
1222
+ "worst_client": 0.8787878787878788,
1223
+ "mean_client": 0.9431776949363156,
1224
+ "comm_MB": 6.53592,
1225
+ "uploads": 278
1226
+ },
1227
+ {
1228
+ "alpha": 0.5,
1229
+ "p_drop": 0.3,
1230
+ "seed": 3,
1231
+ "method": "fedprox",
1232
+ "global_acc": 0.9583333333333334,
1233
+ "worst_client": 0.8787878787878788,
1234
+ "mean_client": 0.9431776949363156,
1235
+ "comm_MB": 6.53592,
1236
+ "uploads": 278
1237
+ },
1238
+ {
1239
+ "alpha": 0.5,
1240
+ "p_drop": 0.3,
1241
+ "seed": 3,
1242
+ "method": "local",
1243
+ "global_acc": 0.6252777777777777,
1244
+ "worst_client": 0.6428571428571429,
1245
+ "mean_client": 0.8898212843385258,
1246
+ "comm_MB": 0.0,
1247
+ "uploads": 0
1248
+ },
1249
+ {
1250
+ "alpha": 0.5,
1251
+ "p_drop": 0.3,
1252
+ "seed": 4,
1253
+ "method": "fedavg",
1254
+ "global_acc": 0.95,
1255
+ "worst_client": 0.875,
1256
+ "mean_client": 0.9595695970695971,
1257
+ "comm_MB": 6.4588,
1258
+ "uploads": 270
1259
+ },
1260
+ {
1261
+ "alpha": 0.5,
1262
+ "p_drop": 0.3,
1263
+ "seed": 4,
1264
+ "method": "fedprox",
1265
+ "global_acc": 0.95,
1266
+ "worst_client": 0.875,
1267
+ "mean_client": 0.9595695970695971,
1268
+ "comm_MB": 6.4588,
1269
+ "uploads": 270
1270
+ },
1271
+ {
1272
+ "alpha": 0.5,
1273
+ "p_drop": 0.3,
1274
+ "seed": 4,
1275
+ "method": "local",
1276
+ "global_acc": 0.6819444444444445,
1277
+ "worst_client": 0.7857142857142857,
1278
+ "mean_client": 0.8983339856501621,
1279
+ "comm_MB": 0.0,
1280
+ "uploads": 0
1281
+ },
1282
+ {
1283
+ "alpha": 5.0,
1284
+ "p_drop": 0.0,
1285
+ "seed": 0,
1286
+ "method": "fedavg",
1287
+ "global_acc": 0.9611111111111111,
1288
+ "worst_client": 0.875,
1289
+ "mean_client": 0.9469440784658175,
1290
+ "comm_MB": 7.712,
1291
+ "uploads": 400
1292
+ },
1293
+ {
1294
+ "alpha": 5.0,
1295
+ "p_drop": 0.0,
1296
+ "seed": 0,
1297
+ "method": "fedprox",
1298
+ "global_acc": 0.9611111111111111,
1299
+ "worst_client": 0.875,
1300
+ "mean_client": 0.9469440784658175,
1301
+ "comm_MB": 7.712,
1302
+ "uploads": 400
1303
+ },
1304
+ {
1305
+ "alpha": 5.0,
1306
+ "p_drop": 0.0,
1307
+ "seed": 0,
1308
+ "method": "local",
1309
+ "global_acc": 0.8724999999999999,
1310
+ "worst_client": 0.8148148148148148,
1311
+ "mean_client": 0.8844483151642072,
1312
+ "comm_MB": 0.0,
1313
+ "uploads": 0
1314
+ },
1315
+ {
1316
+ "alpha": 5.0,
1317
+ "p_drop": 0.0,
1318
+ "seed": 1,
1319
+ "method": "fedavg",
1320
+ "global_acc": 0.9583333333333334,
1321
+ "worst_client": 0.8947368421052632,
1322
+ "mean_client": 0.9690798370164764,
1323
+ "comm_MB": 7.712,
1324
+ "uploads": 400
1325
+ },
1326
+ {
1327
+ "alpha": 5.0,
1328
+ "p_drop": 0.0,
1329
+ "seed": 1,
1330
+ "method": "fedprox",
1331
+ "global_acc": 0.9583333333333334,
1332
+ "worst_client": 0.8947368421052632,
1333
+ "mean_client": 0.9690798370164764,
1334
+ "comm_MB": 7.712,
1335
+ "uploads": 400
1336
+ },
1337
+ {
1338
+ "alpha": 5.0,
1339
+ "p_drop": 0.0,
1340
+ "seed": 1,
1341
+ "method": "local",
1342
+ "global_acc": 0.8722222222222221,
1343
+ "worst_client": 0.7368421052631579,
1344
+ "mean_client": 0.8580957574266623,
1345
+ "comm_MB": 0.0,
1346
+ "uploads": 0
1347
+ },
1348
+ {
1349
+ "alpha": 5.0,
1350
+ "p_drop": 0.0,
1351
+ "seed": 2,
1352
+ "method": "fedavg",
1353
+ "global_acc": 0.9666666666666667,
1354
+ "worst_client": 0.8947368421052632,
1355
+ "mean_client": 0.9676245188831002,
1356
+ "comm_MB": 7.712,
1357
+ "uploads": 400
1358
+ },
1359
+ {
1360
+ "alpha": 5.0,
1361
+ "p_drop": 0.0,
1362
+ "seed": 2,
1363
+ "method": "fedprox",
1364
+ "global_acc": 0.9666666666666667,
1365
+ "worst_client": 0.8947368421052632,
1366
+ "mean_client": 0.9676245188831002,
1367
+ "comm_MB": 7.712,
1368
+ "uploads": 400
1369
+ },
1370
+ {
1371
+ "alpha": 5.0,
1372
+ "p_drop": 0.0,
1373
+ "seed": 2,
1374
+ "method": "local",
1375
+ "global_acc": 0.865,
1376
+ "worst_client": 0.6190476190476191,
1377
+ "mean_client": 0.8972266808536833,
1378
+ "comm_MB": 0.0,
1379
+ "uploads": 0
1380
+ },
1381
+ {
1382
+ "alpha": 5.0,
1383
+ "p_drop": 0.0,
1384
+ "seed": 3,
1385
+ "method": "fedavg",
1386
+ "global_acc": 0.9555555555555556,
1387
+ "worst_client": 0.8620689655172413,
1388
+ "mean_client": 0.9492250490916158,
1389
+ "comm_MB": 7.712,
1390
+ "uploads": 400
1391
+ },
1392
+ {
1393
+ "alpha": 5.0,
1394
+ "p_drop": 0.0,
1395
+ "seed": 3,
1396
+ "method": "fedprox",
1397
+ "global_acc": 0.9527777777777777,
1398
+ "worst_client": 0.8620689655172413,
1399
+ "mean_client": 0.9452250490916159,
1400
+ "comm_MB": 7.712,
1401
+ "uploads": 400
1402
+ },
1403
+ {
1404
+ "alpha": 5.0,
1405
+ "p_drop": 0.0,
1406
+ "seed": 3,
1407
+ "method": "local",
1408
+ "global_acc": 0.8641666666666667,
1409
+ "worst_client": 0.8148148148148148,
1410
+ "mean_client": 0.8797578887324015,
1411
+ "comm_MB": 0.0,
1412
+ "uploads": 0
1413
+ },
1414
+ {
1415
+ "alpha": 5.0,
1416
+ "p_drop": 0.0,
1417
+ "seed": 4,
1418
+ "method": "fedavg",
1419
+ "global_acc": 0.9638888888888889,
1420
+ "worst_client": 0.875,
1421
+ "mean_client": 0.9477089504698201,
1422
+ "comm_MB": 7.712,
1423
+ "uploads": 400
1424
+ },
1425
+ {
1426
+ "alpha": 5.0,
1427
+ "p_drop": 0.0,
1428
+ "seed": 4,
1429
+ "method": "fedprox",
1430
+ "global_acc": 0.9611111111111111,
1431
+ "worst_client": 0.875,
1432
+ "mean_client": 0.9477089504698201,
1433
+ "comm_MB": 7.712,
1434
+ "uploads": 400
1435
+ },
1436
+ {
1437
+ "alpha": 5.0,
1438
+ "p_drop": 0.0,
1439
+ "seed": 4,
1440
+ "method": "local",
1441
+ "global_acc": 0.8794444444444445,
1442
+ "worst_client": 0.7916666666666666,
1443
+ "mean_client": 0.8973321388756172,
1444
+ "comm_MB": 0.0,
1445
+ "uploads": 0
1446
+ },
1447
+ {
1448
+ "alpha": 5.0,
1449
+ "p_drop": 0.3,
1450
+ "seed": 0,
1451
+ "method": "fedavg",
1452
+ "global_acc": 0.95,
1453
+ "worst_client": 0.875,
1454
+ "mean_client": 0.9432403747621139,
1455
+ "comm_MB": 6.6034,
1456
+ "uploads": 285
1457
+ },
1458
+ {
1459
+ "alpha": 5.0,
1460
+ "p_drop": 0.3,
1461
+ "seed": 0,
1462
+ "method": "fedprox",
1463
+ "global_acc": 0.95,
1464
+ "worst_client": 0.875,
1465
+ "mean_client": 0.9432403747621139,
1466
+ "comm_MB": 6.6034,
1467
+ "uploads": 285
1468
+ },
1469
+ {
1470
+ "alpha": 5.0,
1471
+ "p_drop": 0.3,
1472
+ "seed": 0,
1473
+ "method": "local",
1474
+ "global_acc": 0.8724999999999999,
1475
+ "worst_client": 0.8148148148148148,
1476
+ "mean_client": 0.8844483151642072,
1477
+ "comm_MB": 0.0,
1478
+ "uploads": 0
1479
+ },
1480
+ {
1481
+ "alpha": 5.0,
1482
+ "p_drop": 0.3,
1483
+ "seed": 1,
1484
+ "method": "fedavg",
1485
+ "global_acc": 0.9611111111111111,
1486
+ "worst_client": 0.8947368421052632,
1487
+ "mean_client": 0.9647320109295198,
1488
+ "comm_MB": 6.51664,
1489
+ "uploads": 276
1490
+ },
1491
+ {
1492
+ "alpha": 5.0,
1493
+ "p_drop": 0.3,
1494
+ "seed": 1,
1495
+ "method": "fedprox",
1496
+ "global_acc": 0.9611111111111111,
1497
+ "worst_client": 0.8947368421052632,
1498
+ "mean_client": 0.9647320109295198,
1499
+ "comm_MB": 6.51664,
1500
+ "uploads": 276
1501
+ },
1502
+ {
1503
+ "alpha": 5.0,
1504
+ "p_drop": 0.3,
1505
+ "seed": 1,
1506
+ "method": "local",
1507
+ "global_acc": 0.8722222222222221,
1508
+ "worst_client": 0.7368421052631579,
1509
+ "mean_client": 0.8580957574266623,
1510
+ "comm_MB": 0.0,
1511
+ "uploads": 0
1512
+ },
1513
+ {
1514
+ "alpha": 5.0,
1515
+ "p_drop": 0.3,
1516
+ "seed": 2,
1517
+ "method": "fedavg",
1518
+ "global_acc": 0.9638888888888889,
1519
+ "worst_client": 0.9090909090909091,
1520
+ "mean_client": 0.9641755555657158,
1521
+ "comm_MB": 6.53592,
1522
+ "uploads": 278
1523
+ },
1524
+ {
1525
+ "alpha": 5.0,
1526
+ "p_drop": 0.3,
1527
+ "seed": 2,
1528
+ "method": "fedprox",
1529
+ "global_acc": 0.9611111111111111,
1530
+ "worst_client": 0.9090909090909091,
1531
+ "mean_client": 0.9641755555657158,
1532
+ "comm_MB": 6.53592,
1533
+ "uploads": 278
1534
+ },
1535
+ {
1536
+ "alpha": 5.0,
1537
+ "p_drop": 0.3,
1538
+ "seed": 2,
1539
+ "method": "local",
1540
+ "global_acc": 0.865,
1541
+ "worst_client": 0.6190476190476191,
1542
+ "mean_client": 0.8972266808536833,
1543
+ "comm_MB": 0.0,
1544
+ "uploads": 0
1545
+ },
1546
+ {
1547
+ "alpha": 5.0,
1548
+ "p_drop": 0.3,
1549
+ "seed": 3,
1550
+ "method": "fedavg",
1551
+ "global_acc": 0.9611111111111111,
1552
+ "worst_client": 0.896551724137931,
1553
+ "mean_client": 0.9603770286573885,
1554
+ "comm_MB": 6.53592,
1555
+ "uploads": 278
1556
+ },
1557
+ {
1558
+ "alpha": 5.0,
1559
+ "p_drop": 0.3,
1560
+ "seed": 3,
1561
+ "method": "fedprox",
1562
+ "global_acc": 0.9611111111111111,
1563
+ "worst_client": 0.896551724137931,
1564
+ "mean_client": 0.9603770286573885,
1565
+ "comm_MB": 6.53592,
1566
+ "uploads": 278
1567
+ },
1568
+ {
1569
+ "alpha": 5.0,
1570
+ "p_drop": 0.3,
1571
+ "seed": 3,
1572
+ "method": "local",
1573
+ "global_acc": 0.8641666666666667,
1574
+ "worst_client": 0.8148148148148148,
1575
+ "mean_client": 0.8797578887324015,
1576
+ "comm_MB": 0.0,
1577
+ "uploads": 0
1578
+ },
1579
+ {
1580
+ "alpha": 5.0,
1581
+ "p_drop": 0.3,
1582
+ "seed": 4,
1583
+ "method": "fedavg",
1584
+ "global_acc": 0.9638888888888889,
1585
+ "worst_client": 0.875,
1586
+ "mean_client": 0.9435422838031535,
1587
+ "comm_MB": 6.4588,
1588
+ "uploads": 270
1589
+ },
1590
+ {
1591
+ "alpha": 5.0,
1592
+ "p_drop": 0.3,
1593
+ "seed": 4,
1594
+ "method": "fedprox",
1595
+ "global_acc": 0.9638888888888889,
1596
+ "worst_client": 0.875,
1597
+ "mean_client": 0.9435422838031535,
1598
+ "comm_MB": 6.4588,
1599
+ "uploads": 270
1600
+ },
1601
+ {
1602
+ "alpha": 5.0,
1603
+ "p_drop": 0.3,
1604
+ "seed": 4,
1605
+ "method": "local",
1606
+ "global_acc": 0.8794444444444445,
1607
+ "worst_client": 0.7916666666666666,
1608
+ "mean_client": 0.8973321388756172,
1609
+ "comm_MB": 0.0,
1610
+ "uploads": 0
1611
+ }
1612
+ ],
1613
+ "tests": {
1614
+ "reload": {
1615
+ "returncode": 0,
1616
+ "seconds": 0.19,
1617
+ "process": "fresh python -E subprocess, cwd = temporary copy of the model folder",
1618
+ "passed": true,
1619
+ "max_abs_diff": 2.2687930867348882e-08,
1620
+ "sim_curve_diff": 0.0,
1621
+ "n_samples": 16
1622
+ },
1623
+ "node_parity": {
1624
+ "cases": 4,
1625
+ "failed": 0,
1626
+ "maxDiff": 0,
1627
+ "results": [
1628
+ {
1629
+ "name": "fedavg-a0.3-p0.2",
1630
+ "maxDiff": 0
1631
+ },
1632
+ {
1633
+ "name": "fedprox-a0.1-p0.0",
1634
+ "maxDiff": 0
1635
+ },
1636
+ {
1637
+ "name": "local-a2.0-p0.0",
1638
+ "maxDiff": 0
1639
+ },
1640
+ {
1641
+ "name": "fedavg-a5.0-p0.5",
1642
+ "maxDiff": 0
1643
+ }
1644
+ ],
1645
+ "ms": 585,
1646
+ "runtime": "node v22.23.4",
1647
+ "status": "passed"
1648
+ },
1649
+ "browser": {
1650
+ "status": "skipped",
1651
+ "reason": "playwright not installed"
1652
+ }
1653
+ }
1654
+ }
fed_core.py ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Next Federation: simulated federated learning on clients with unequal data.
2
+
3
+ Simulation only: all clients run in one process. Mirrored line by line in web/core.js.
4
+ Model: MLP 64 -> 32 (ReLU) -> 10, softmax cross-entropy, minibatch SGD.
5
+ Methods:
6
+ fedavg - clients run E local epochs from the global model; server averages uploads weighted by sample count
7
+ fedprox - same, plus proximal term (mu/2)||w - w_global||^2 in the local objective
8
+ local - every client trains alone for the same number of local epochs in total (no communication)
9
+ Client dropout: each round each client independently fails to return its update with probability p_drop.
10
+ """
11
+ import math
12
+
13
+ import numpy as np
14
+
15
+ from nlrt import Mulberry32
16
+
17
+ D_IN, D_HID, D_OUT = 64, 32, 10
18
+ N_PARAMS = D_IN * D_HID + D_HID + D_HID * D_OUT + D_OUT
19
+
20
+
21
+ def gamma_sample(rng, a):
22
+ """Marsaglia-Tsang gamma(a, 1); boosted for a < 1."""
23
+ if a < 1:
24
+ u = max(rng.random(), 1e-12)
25
+ return gamma_sample(rng, a + 1) * u ** (1.0 / a)
26
+ d = a - 1.0 / 3
27
+ c = 1.0 / math.sqrt(9 * d)
28
+ while True:
29
+ x = rng.normal()
30
+ v = 1 + c * x
31
+ if v <= 0:
32
+ continue
33
+ v = v * v * v
34
+ u = rng.random()
35
+ if u < 1 - 0.0331 * (x * x) * (x * x) or math.log(max(u, 1e-300)) < 0.5 * x * x + d * (1 - v + math.log(v)):
36
+ return d * v
37
+
38
+
39
+ def partition(y, n_clients, alpha, seed):
40
+ """Dirichlet label skew: for each class, split its examples over clients with Dir(alpha) proportions."""
41
+ rng = Mulberry32(seed * 977 + 13)
42
+ clients = [[] for _ in range(n_clients)]
43
+ for c in range(10):
44
+ idx = [i for i in range(len(y)) if y[i] == c]
45
+ for k in range(len(idx) - 1, 0, -1): # Fisher-Yates
46
+ j = rng.randint(k + 1)
47
+ idx[k], idx[j] = idx[j], idx[k]
48
+ g = [gamma_sample(rng, alpha) for _ in range(n_clients)]
49
+ s = sum(g)
50
+ cuts, acc = [], 0.0
51
+ for k in range(n_clients - 1):
52
+ acc += g[k] / s
53
+ cuts.append(int(math.floor(acc * len(idx) + 0.5)))
54
+ lo = 0
55
+ for k in range(n_clients):
56
+ hi = cuts[k] if k < n_clients - 1 else len(idx)
57
+ clients[k].extend(idx[lo:hi])
58
+ lo = hi
59
+ out = []
60
+ for k in range(n_clients):
61
+ ids = sorted(clients[k])
62
+ for i in range(len(ids) - 1, 0, -1):
63
+ j = rng.randint(i + 1)
64
+ ids[i], ids[j] = ids[j], ids[i]
65
+ n_test = max(1, len(ids) // 5) if len(ids) >= 5 else 0
66
+ out.append({"train": ids[n_test:], "test": ids[:n_test]})
67
+ return out
68
+
69
+
70
+ def init_params(seed):
71
+ rng = Mulberry32(seed * 131 + 5)
72
+ w1 = np.array([[rng.normal() * math.sqrt(2.0 / D_IN) for _ in range(D_IN)] for _ in range(D_HID)])
73
+ w2 = np.array([[rng.normal() * math.sqrt(1.0 / D_HID) for _ in range(D_HID)] for _ in range(D_OUT)])
74
+ return {"w1": w1, "b1": np.zeros(D_HID), "w2": w2, "b2": np.zeros(D_OUT)}
75
+
76
+
77
+ def copy(p):
78
+ return {k: v.copy() for k, v in p.items()}
79
+
80
+
81
+ def forward(p, X):
82
+ h = np.maximum(X @ p["w1"].T + p["b1"], 0)
83
+ z = h @ p["w2"].T + p["b2"]
84
+ return h, z
85
+
86
+
87
+ def predict(p, X):
88
+ return forward(p, X)[1].argmax(1)
89
+
90
+
91
+ def accuracy(p, X, y):
92
+ if len(y) == 0:
93
+ return -1.0 # sentinel: client has no local test data
94
+ return float((predict(p, X) == y).mean())
95
+
96
+
97
+ def local_train(p, X, y, epochs, lr, batch, rng, mu=0.0, anchor=None):
98
+ n = len(y)
99
+ order = list(range(n))
100
+ for _ in range(epochs):
101
+ for k in range(n - 1, 0, -1):
102
+ j = rng.randint(k + 1)
103
+ order[k], order[j] = order[j], order[k]
104
+ for s in range(0, n, batch):
105
+ ids = order[s:s + batch]
106
+ xb, yb = X[ids], y[ids]
107
+ h, z = forward(p, xb)
108
+ z = z - z.max(axis=1, keepdims=True)
109
+ e = np.exp(z)
110
+ P = e / e.sum(axis=1, keepdims=True)
111
+ P[np.arange(len(ids)), yb] -= 1.0
112
+ P /= len(ids)
113
+ g2 = P.T @ h
114
+ gb2 = P.sum(axis=0)
115
+ dh = (P @ p["w2"]) * (h > 0)
116
+ g1 = dh.T @ xb
117
+ gb1 = dh.sum(axis=0)
118
+ grads = {"w1": g1, "b1": gb1, "w2": g2, "b2": gb2}
119
+ for key in ("w1", "b1", "w2", "b2"):
120
+ g = grads[key]
121
+ if mu > 0:
122
+ g = g + mu * (p[key] - anchor[key])
123
+ p[key] = p[key] - lr * g
124
+ return p
125
+
126
+
127
+ def run(X, y, parts, Xg, yg, method, rounds=30, epochs=2, lr=0.1, batch=20, mu=0.1, p_drop=0.0, seed=0):
128
+ """Returns per-round global accuracy, final per-client accuracies and communication bytes."""
129
+ rng = Mulberry32(seed * 7907 + 11)
130
+ drop_rng = Mulberry32(seed * 104729 + 3)
131
+ nc = len(parts)
132
+ g = init_params(seed)
133
+ curve, comm_bytes, uploads = [], 0, 0
134
+ if method == "local":
135
+ models = [copy(g) for _ in range(nc)]
136
+ for r in range(rounds):
137
+ for k in range(nc):
138
+ tr = parts[k]["train"]
139
+ if tr:
140
+ local_train(models[k], X[tr], y[tr], epochs, lr, batch, rng)
141
+ curve.append(float(np.mean([accuracy(models[k], Xg, yg) for k in range(nc)])))
142
+ client_acc = [accuracy(models[k], X[parts[k]["test"]], y[parts[k]["test"]]) for k in range(nc)]
143
+ return {"curve": curve, "client_acc": client_acc, "comm_bytes": 0, "uploads": 0, "final": models}
144
+ for r in range(rounds):
145
+ updates = []
146
+ for k in range(nc):
147
+ comm_bytes += N_PARAMS * 4 # download of the global model
148
+ tr = parts[k]["train"]
149
+ dropped = drop_rng.random() < p_drop
150
+ if not tr:
151
+ continue
152
+ local = local_train(copy(g), X[tr], y[tr], epochs, lr, batch, rng,
153
+ mu=mu if method == "fedprox" else 0.0, anchor=g)
154
+ if dropped:
155
+ continue
156
+ comm_bytes += N_PARAMS * 4
157
+ uploads += 1
158
+ updates.append((len(tr), local))
159
+ if updates:
160
+ total = float(sum(n for n, _ in updates))
161
+ g = {key: sum((n / total) * u[key] for n, u in updates) for key in g}
162
+ curve.append(accuracy(g, Xg, yg))
163
+ client_acc = [accuracy(g, X[parts[k]["test"]], y[parts[k]["test"]]) for k in range(nc)]
164
+ return {"curve": curve, "client_acc": client_acc, "comm_bytes": comm_bytes, "uploads": uploads, "final": g}
inference.py ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Next Federation: load the released global model (FedProx) or rerun the federated simulation.
2
+
3
+ Usage:
4
+ python inference.py --selftest
5
+ from inference import load, predict
6
+ """
7
+ import json
8
+ import os
9
+ import sys
10
+
11
+ HERE = os.path.dirname(os.path.abspath(__file__))
12
+ sys.path.insert(0, HERE)
13
+
14
+ import numpy as np # noqa: E402
15
+
16
+ import fed_core as fc # noqa: E402
17
+ from nlrt import forward, load_model # noqa: E402
18
+
19
+
20
+ def load(path=os.path.join(HERE, "model.json")):
21
+ return load_model(path)
22
+
23
+
24
+ def predict(spec, pixels):
25
+ """pixels: (N, 64) ints 0..16. Returns class probabilities (N, 10) from the global model."""
26
+ z = forward(spec["nets"]["global"], np.asarray(pixels, dtype=float) / 16.0)
27
+ z = z - z.max(axis=1, keepdims=True)
28
+ e = np.exp(z)
29
+ return e / e.sum(axis=1, keepdims=True)
30
+
31
+
32
+ def _selftest():
33
+ spec = load()
34
+ with open(os.path.join(HERE, "selftest.json"), encoding="utf-8") as f:
35
+ t = json.load(f)
36
+ P = predict(spec, t["pixels"])
37
+ diff = float(np.abs(P - np.asarray(t["expected_proba"])).max())
38
+ # rerun a tiny federated simulation and compare with stored output
39
+ X = np.asarray(t["sim"]["pixels"], dtype=float) / 16.0
40
+ y = np.asarray(t["sim"]["y"])
41
+ parts = fc.partition(y.tolist(), 4, 0.5, 2)
42
+ gt = list(range(len(y)))
43
+ r = fc.run(X, y, parts, X[gt], y[gt], "fedprox", rounds=2, seed=2, **spec["meta"]["hyper"])
44
+ sim_diff = float(np.abs(np.asarray(r["curve"]) - np.asarray(t["sim"]["expected_curve"])).max())
45
+ ok = diff < 1e-6 and sim_diff < 1e-9
46
+ print(json.dumps({"passed": ok, "max_abs_diff": diff, "sim_curve_diff": sim_diff, "n_samples": len(t["pixels"])}))
47
+ return ok
48
+
49
+
50
+ if __name__ == "__main__":
51
+ if "--selftest" in sys.argv:
52
+ sys.exit(0 if _selftest() else 1)
53
+ print(__doc__)
model.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"format":"nlrt-1","meta":{"name":"Next Federation","hyper":{"epochs":2,"lr":0.3,"batch":20,"mu":0.01},"architecture":"MLP 64-32-10 (ReLU)","input":"8x8 digit pixels / 16","released_run":"FedProx, alpha=0.5, dropout=0.3, seed 0, 40 rounds"},"nets":{"global":[{"type":"linear","w":"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","b":"K2F3vUv5H7yhoqA9i8Z7PQAAAACM0OQ89XKKPcFEWD3Rjxs+7YovPgAAAADoJOK80PZePfRXvD3YW+c9bk11vOSPGT4mvAu8kca7PQnETz4XzR4+FjUfvWkfjzwK4/U90nr0PfMkF7yIg9U9A6IVPoVMMT06Bsk8svE3PicU8Tw=","w_shape":[32,64],"b_shape":[32]},{"type":"relu"},{"type":"linear","w":"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","b":"j+5dvanAn73cDQq+FF4fPt54ar2IZxA+fZMGva+1oD3I63k8FEkIvQ==","w_shape":[10,32],"b_shape":[10]}]},"tensors":{}}
nlrt.py ADDED
@@ -0,0 +1,226 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """nlrt: tiny portable runtime shared by Next Lab models.
2
+
3
+ The same layer list runs in NumPy (this file) and in the browser (nlrt.js).
4
+ Weights are stored as base64 little-endian float32 inside model.json.
5
+ Only NumPy is required to load and run a model.
6
+ """
7
+ import base64
8
+ import json
9
+ import math
10
+
11
+ import numpy as np
12
+
13
+ __all__ = ["b64e", "b64d", "forward", "load_model", "Mulberry32", "export_sequential"]
14
+
15
+
16
+ def b64e(a):
17
+ return base64.b64encode(np.ascontiguousarray(a, dtype="<f4").tobytes()).decode("ascii")
18
+
19
+
20
+ def b64d(s, shape=None):
21
+ a = np.frombuffer(base64.b64decode(s), dtype="<f4").astype(np.float64)
22
+ return a.reshape(shape) if shape is not None else a
23
+
24
+
25
+ def load_model(path):
26
+ with open(path, "r", encoding="utf-8") as f:
27
+ spec = json.load(f)
28
+ for net in spec.get("nets", {}).values():
29
+ _decode(net)
30
+ for k, v in list(spec.get("tensors", {}).items()):
31
+ spec["tensors"][k] = b64d(v["data"], v["shape"])
32
+ return spec
33
+
34
+
35
+ def _decode(layers):
36
+ for L in layers:
37
+ for key in ("w", "b"):
38
+ if key in L and isinstance(L[key], str):
39
+ L[key] = b64d(L[key], L.get(key + "_shape"))
40
+
41
+
42
+ class Mulberry32:
43
+ """Deterministic PRNG with an identical JavaScript twin in nlrt.js."""
44
+
45
+ def __init__(self, seed):
46
+ self.a = seed & 0xFFFFFFFF
47
+
48
+ def random(self):
49
+ self.a = (self.a + 0x6D2B79F5) & 0xFFFFFFFF
50
+ t = self.a
51
+ t = ((t ^ (t >> 15)) * (t | 1)) & 0xFFFFFFFF
52
+ t ^= (t + (((t ^ (t >> 7)) * (t | 61)) & 0xFFFFFFFF)) & 0xFFFFFFFF
53
+ return ((t ^ (t >> 14)) & 0xFFFFFFFF) / 4294967296.0
54
+
55
+ def uniform(self, lo, hi):
56
+ return lo + (hi - lo) * self.random()
57
+
58
+ def randint(self, n):
59
+ return int(math.floor(self.random() * n))
60
+
61
+ def normal(self):
62
+ u = max(self.random(), 1e-12)
63
+ v = self.random()
64
+ return math.sqrt(-2.0 * math.log(u)) * math.cos(2.0 * math.pi * v)
65
+
66
+
67
+ def _pair(v):
68
+ return list(v) if isinstance(v, (list, tuple)) else [v, v]
69
+
70
+
71
+ def _conv2d(x, w, b, stride, pad):
72
+ n, c, h, wd = x.shape
73
+ o, _, kh, kw = w.shape
74
+ sh, sw = stride
75
+ ph, pw = pad
76
+ xp = np.pad(x, ((0, 0), (0, 0), (ph, ph), (pw, pw)))
77
+ oh = (h + 2 * ph - kh) // sh + 1
78
+ ow = (wd + 2 * pw - kw) // sw + 1
79
+ cols = np.empty((n, c, kh, kw, oh, ow), dtype=x.dtype)
80
+ for i in range(kh):
81
+ for j in range(kw):
82
+ cols[:, :, i, j] = xp[:, :, i:i + sh * oh:sh, j:j + sw * ow:sw]
83
+ out = np.tensordot(cols, w, axes=([1, 2, 3], [1, 2, 3])) # n, oh, ow, o
84
+ out = out.transpose(0, 3, 1, 2)
85
+ if b is not None:
86
+ out = out + b[None, :, None, None]
87
+ return out
88
+
89
+
90
+ def _maxpool(x, k):
91
+ kh, kw = k
92
+ n, c, h, w = x.shape
93
+ oh, ow = h // kh, w // kw
94
+ x = x[:, :, :oh * kh, :ow * kw].reshape(n, c, oh, kh, ow, kw)
95
+ return x.max(axis=(3, 5))
96
+
97
+
98
+ def forward(layers, x):
99
+ """Run a layer list on a batch. x: (N, D) or (N, C, H, W) or (N, C, L)."""
100
+ x = np.asarray(x, dtype=np.float64)
101
+ for L in layers:
102
+ t = L["type"]
103
+ if t == "linear":
104
+ x = x @ L["w"].T + L["b"]
105
+ elif t == "conv2d":
106
+ squeeze = x.ndim == 3
107
+ if squeeze:
108
+ x = x[:, :, None, :]
109
+ x = _conv2d(x, L["w"], L["b"], _pair(L["stride"]), _pair(L["pad"]))
110
+ if squeeze:
111
+ x = x[:, :, 0, :]
112
+ elif t == "relu":
113
+ x = np.maximum(x, 0)
114
+ elif t == "tanh":
115
+ x = np.tanh(x)
116
+ elif t == "sigmoid":
117
+ x = 1 / (1 + np.exp(-x))
118
+ elif t == "maxpool":
119
+ squeeze = x.ndim == 3
120
+ if squeeze:
121
+ x = x[:, :, None, :]
122
+ x = _maxpool(x, _pair(L["k"]))
123
+ if squeeze:
124
+ x = x[:, :, 0, :]
125
+ elif t == "gap":
126
+ x = x.reshape(x.shape[0], x.shape[1], -1).mean(axis=2)
127
+ elif t == "flatten":
128
+ x = x.reshape(x.shape[0], -1)
129
+ elif t == "upsample":
130
+ x = x.repeat(L["s"], axis=-1)
131
+ if x.ndim == 4:
132
+ x = x.repeat(L["s"], axis=-2)
133
+ elif t == "l2norm":
134
+ x = x / np.maximum(np.linalg.norm(x, axis=1, keepdims=True), 1e-12)
135
+ elif t == "reshape":
136
+ x = x.reshape([x.shape[0]] + list(L["shape"]))
137
+ else:
138
+ raise ValueError("unknown layer " + t)
139
+ return x
140
+
141
+
142
+ def export_sequential(seq):
143
+ """Convert a torch nn.Sequential into a portable layer list (BatchNorm folded)."""
144
+ import torch.nn as nn
145
+
146
+ mods = list(seq.modules())[1:] if not isinstance(seq, (list, tuple)) else list(seq)
147
+ mods = [m for m in mods if not isinstance(m, nn.Sequential)]
148
+ layers = []
149
+
150
+ def tensor(t):
151
+ return t.detach().cpu().double().numpy()
152
+
153
+ for m in mods:
154
+ if isinstance(m, nn.Linear):
155
+ b = tensor(m.bias) if m.bias is not None else np.zeros(m.out_features)
156
+ layers.append({"type": "linear", "w": tensor(m.weight), "b": b})
157
+ elif isinstance(m, (nn.Conv2d, nn.Conv1d)):
158
+ w = tensor(m.weight)
159
+ b = tensor(m.bias) if m.bias is not None else np.zeros(w.shape[0])
160
+ if isinstance(m, nn.Conv1d):
161
+ w = w[:, :, None, :]
162
+ stride, pad = [1, m.stride[0]], [0, m.padding[0]]
163
+ else:
164
+ stride, pad = list(m.stride), list(m.padding)
165
+ layers.append({"type": "conv2d", "w": w, "b": b, "stride": stride, "pad": pad})
166
+ elif isinstance(m, (nn.BatchNorm1d, nn.BatchNorm2d)):
167
+ prev = layers[-1]
168
+ assert prev["type"] in ("conv2d", "linear"), "BatchNorm must follow conv/linear"
169
+ scale = tensor(m.weight) / np.sqrt(tensor(m.running_var) + m.eps)
170
+ shift = tensor(m.bias) - tensor(m.running_mean) * scale
171
+ if prev["type"] == "conv2d":
172
+ prev["w"] = prev["w"] * scale[:, None, None, None]
173
+ else:
174
+ prev["w"] = prev["w"] * scale[:, None]
175
+ prev["b"] = prev["b"] * scale + shift
176
+ elif isinstance(m, nn.ReLU):
177
+ layers.append({"type": "relu"})
178
+ elif isinstance(m, nn.Tanh):
179
+ layers.append({"type": "tanh"})
180
+ elif isinstance(m, nn.Sigmoid):
181
+ layers.append({"type": "sigmoid"})
182
+ elif isinstance(m, nn.MaxPool2d):
183
+ layers.append({"type": "maxpool", "k": _pair(m.kernel_size)})
184
+ elif isinstance(m, nn.MaxPool1d):
185
+ k = m.kernel_size if isinstance(m, int) else m.kernel_size
186
+ layers.append({"type": "maxpool", "k": [1, k if isinstance(k, int) else k[0]]})
187
+ elif isinstance(m, (nn.AdaptiveAvgPool2d, nn.AdaptiveAvgPool1d)):
188
+ layers.append({"type": "gap"})
189
+ elif isinstance(m, nn.Flatten):
190
+ layers.append({"type": "flatten"})
191
+ elif isinstance(m, nn.Upsample):
192
+ layers.append({"type": "upsample", "s": int(m.scale_factor)})
193
+ elif isinstance(m, nn.Dropout):
194
+ continue
195
+ elif m.__class__.__name__ == "L2Norm":
196
+ layers.append({"type": "l2norm"})
197
+ elif m.__class__.__name__ == "Reshape":
198
+ layers.append({"type": "reshape", "shape": list(m.shape)})
199
+ else:
200
+ raise ValueError("cannot export " + m.__class__.__name__)
201
+ return layers
202
+
203
+
204
+ def encode_layers(layers):
205
+ out = []
206
+ for L in layers:
207
+ L2 = dict(L)
208
+ for key in ("w", "b"):
209
+ if key in L2:
210
+ arr = np.asarray(L2[key])
211
+ L2[key + "_shape"] = list(arr.shape)
212
+ L2[key] = b64e(arr)
213
+ out.append(L2)
214
+ return out
215
+
216
+
217
+ def save_model(path, nets=None, tensors=None, meta=None):
218
+ spec = {"format": "nlrt-1", "meta": meta or {}, "nets": {}, "tensors": {}}
219
+ for name, layers in (nets or {}).items():
220
+ spec["nets"][name] = encode_layers(layers)
221
+ for name, arr in (tensors or {}).items():
222
+ arr = np.asarray(arr)
223
+ spec["tensors"][name] = {"shape": list(arr.shape), "data": b64e(arr)}
224
+ with open(path, "w", encoding="utf-8") as f:
225
+ json.dump(spec, f, separators=(",", ":"))
226
+ return spec
preview.png ADDED
selftest.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"pixels": [[0, 0, 0, 12, 13, 5, 0, 0, 0, 0, 0, 11, 16, 9, 0, 0, 0, 0, 3, 15, 16, 6, 0, 0, 0, 7, 15, 16, 16, 2, 0, 0, 0, 0, 1, 16, 16, 3, 0, 0, 0, 0, 1, 16, 16, 6, 0, 0, 0, 0, 1, 16, 16, 6, 0, 0, 0, 0, 0, 11, 16, 10, 0, 0], [0, 0, 7, 15, 13, 1, 0, 0, 0, 8, 13, 6, 15, 4, 0, 0, 0, 2, 1, 13, 13, 0, 0, 0, 0, 0, 2, 15, 11, 1, 0, 0, 0, 0, 0, 1, 12, 12, 1, 0, 0, 0, 0, 0, 1, 10, 8, 0, 0, 0, 8, 4, 5, 14, 9, 0, 0, 0, 7, 13, 13, 9, 0, 0], [0, 0, 0, 12, 13, 0, 0, 0, 0, 0, 5, 16, 8, 0, 0, 0, 0, 0, 13, 16, 3, 0, 0, 0, 0, 0, 14, 13, 0, 0, 0, 0, 0, 0, 15, 12, 7, 2, 0, 0, 0, 0, 13, 16, 13, 16, 3, 0, 0, 0, 7, 16, 11, 15, 8, 0, 0, 0, 1, 9, 15, 11, 3, 0], [0, 0, 11, 12, 0, 0, 0, 0, 0, 2, 16, 16, 16, 13, 0, 0, 0, 3, 16, 12, 10, 14, 0, 0, 0, 1, 16, 1, 12, 15, 0, 0, 0, 0, 13, 16, 9, 15, 2, 0, 0, 0, 0, 3, 0, 9, 11, 0, 0, 0, 0, 0, 9, 15, 4, 0, 0, 0, 9, 12, 13, 3, 0, 0], [0, 2, 9, 15, 14, 9, 3, 0, 0, 4, 13, 8, 9, 16, 8, 0, 0, 0, 0, 6, 14, 15, 3, 0, 0, 0, 0, 11, 14, 2, 0, 0, 0, 0, 0, 2, 15, 11, 0, 0, 0, 0, 0, 0, 2, 15, 4, 0, 0, 1, 5, 6, 13, 16, 6, 0, 0, 2, 12, 12, 13, 11, 0, 0], [0, 0, 0, 8, 15, 1, 0, 0, 0, 0, 1, 14, 13, 1, 1, 0, 0, 0, 10, 15, 3, 15, 11, 0, 0, 7, 16, 7, 1, 16, 8, 0, 0, 9, 16, 13, 14, 16, 5, 0, 0, 1, 10, 15, 16, 14, 0, 0, 0, 0, 0, 1, 16, 10, 0, 0, 0, 0, 0, 10, 15, 4, 0, 0], [0, 5, 12, 13, 16, 16, 2, 0, 0, 11, 16, 15, 8, 4, 0, 0, 0, 8, 14, 11, 1, 0, 0, 0, 0, 8, 16, 16, 14, 0, 0, 0, 0, 1, 6, 6, 16, 0, 0, 0, 0, 0, 0, 5, 16, 3, 0, 0, 0, 1, 5, 15, 13, 0, 0, 0, 0, 4, 15, 16, 2, 0, 0, 0], [0, 0, 6, 14, 4, 0, 0, 0, 0, 0, 11, 16, 10, 0, 0, 0, 0, 0, 8, 14, 16, 2, 0, 0, 0, 0, 1, 12, 12, 11, 0, 0, 0, 0, 0, 0, 0, 11, 3, 0, 0, 0, 0, 0, 0, 5, 11, 0, 0, 0, 1, 4, 4, 7, 16, 2, 0, 0, 7, 16, 16, 13, 11, 1], [0, 0, 8, 16, 5, 0, 0, 0, 0, 1, 13, 11, 16, 0, 0, 0, 0, 0, 10, 0, 13, 3, 0, 0, 0, 0, 3, 1, 16, 1, 0, 0, 0, 0, 0, 9, 12, 0, 0, 0, 0, 0, 3, 15, 5, 0, 0, 0, 0, 0, 14, 15, 8, 8, 3, 0, 0, 0, 7, 12, 12, 12, 13, 1], [0, 0, 0, 0, 12, 2, 0, 0, 0, 0, 0, 6, 14, 1, 0, 0, 0, 0, 4, 16, 7, 8, 0, 0, 0, 0, 13, 9, 0, 16, 6, 0, 0, 6, 16, 10, 11, 16, 0, 0, 0, 0, 5, 10, 13, 16, 0, 0, 0, 0, 0, 0, 6, 16, 0, 0, 0, 0, 0, 0, 12, 8, 0, 0], [0, 0, 12, 8, 8, 7, 0, 0, 0, 3, 16, 16, 11, 7, 0, 0, 0, 2, 14, 1, 0, 0, 0, 0, 0, 5, 14, 5, 0, 0, 0, 0, 0, 2, 15, 16, 9, 0, 0, 0, 0, 0, 0, 2, 16, 2, 0, 0, 0, 0, 4, 8, 16, 4, 0, 0, 0, 0, 11, 14, 9, 0, 0, 0], [0, 0, 9, 13, 7, 0, 0, 0, 0, 0, 12, 16, 16, 2, 0, 0, 0, 0, 12, 13, 16, 6, 0, 0, 0, 0, 6, 16, 16, 14, 0, 0, 0, 0, 0, 0, 2, 16, 3, 0, 0, 0, 0, 0, 0, 9, 10, 0, 0, 0, 3, 7, 12, 14, 16, 2, 0, 0, 7, 12, 12, 12, 11, 0], [0, 0, 2, 13, 8, 0, 0, 0, 0, 0, 6, 16, 16, 6, 0, 0, 0, 0, 5, 15, 13, 11, 0, 0, 0, 0, 0, 7, 16, 15, 0, 0, 0, 0, 0, 0, 0, 14, 3, 0, 0, 0, 0, 0, 0, 7, 11, 0, 0, 0, 0, 3, 4, 4, 16, 2, 0, 0, 2, 15, 13, 14, 13, 2], [0, 0, 7, 7, 13, 16, 4, 0, 0, 0, 13, 13, 6, 12, 7, 0, 0, 0, 10, 4, 10, 11, 1, 0, 0, 0, 8, 16, 10, 0, 0, 0, 0, 3, 14, 16, 0, 0, 0, 0, 0, 8, 8, 11, 5, 0, 0, 0, 0, 4, 10, 9, 8, 0, 0, 0, 0, 1, 11, 16, 6, 0, 0, 0], [0, 0, 9, 16, 13, 6, 0, 0, 0, 0, 6, 5, 16, 16, 0, 0, 0, 0, 0, 8, 15, 5, 0, 0, 0, 0, 0, 5, 14, 3, 0, 0, 0, 0, 0, 0, 9, 15, 2, 0, 0, 0, 0, 0, 0, 11, 12, 0, 0, 0, 4, 8, 11, 15, 12, 0, 0, 0, 11, 14, 12, 8, 0, 0], [0, 0, 4, 8, 16, 5, 0, 0, 0, 0, 9, 16, 8, 11, 0, 0, 0, 0, 5, 10, 0, 13, 2, 0, 0, 0, 0, 13, 4, 15, 2, 0, 0, 0, 0, 9, 16, 8, 0, 0, 0, 0, 8, 15, 14, 5, 0, 0, 0, 0, 16, 5, 14, 4, 0, 0, 0, 0, 6, 16, 12, 1, 0, 0]], "expected_proba": [[7.52761009770872e-09, 0.9967753105133723, 1.879691890173269e-05, 2.777775595905777e-05, 0.0018097638262452517, 2.8098328717405307e-07, 1.463709555451008e-05, 1.6552472848974414e-06, 0.0013474716345934685, 4.298497191533457e-06], [4.567402457782867e-07, 5.326239646048198e-05, 0.00017146222233767057, 0.9976613790883739, 4.596573908122248e-08, 0.0008838141905925958, 3.6888305411480303e-06, 1.1622806916411836e-06, 0.0001703108448654642, 0.0010544174401521656], [6.161074394317075e-05, 0.00020234480845835893, 1.82641033854841e-05, 6.728422210385693e-08, 0.0005144815331655104, 7.559307093938214e-06, 0.9976684500708449, 2.26499620645797e-08, 0.0015271806988121106, 1.8800112381039706e-08], [0.0023457670419469545, 0.0005563907025530685, 2.881437714468756e-05, 0.0004648803885894567, 0.00037286741680773506, 0.005199827549086472, 8.44221127432726e-06, 0.0006911323113914361, 0.003151482225780746, 0.987180395775425], [1.831202767669017e-08, 5.894788509000839e-07, 5.629563165397909e-05, 0.9998138381031161, 1.605863149359371e-10, 7.576687895736433e-06, 2.9536362688228884e-09, 1.4911235002361351e-06, 5.866932302647859e-05, 6.151822570653958e-05], [0.00020050235182333143, 9.229150806526564e-05, 2.4764320673233204e-08, 1.0809172529719882e-09, 0.9993625486916464, 8.529304312154553e-07, 0.0002216007223119834, 2.184725486722245e-05, 9.646607505159248e-05, 3.864620564973943e-06], [4.9952942123718e-07, 1.3940485942766641e-05, 1.592471783917852e-06, 0.00013934296253683066, 1.2880484996732723e-05, 0.9997389188409883, 6.44345800539871e-07, 4.6955452070644306e-05, 2.0834579302061983e-05, 2.4390847156859906e-05], [0.00011269002438740374, 0.013018049130472378, 0.016459944206858073, 0.42784123009131453, 1.9858720166600533e-05, 0.007195663525214936, 0.00033326978154181214, 3.442859062735857e-05, 0.003700611986492232, 0.5312842539429247], [4.358227455366593e-07, 0.00010403340816168912, 0.9991933122772223, 0.00042450007759862156, 1.1330611265925229e-07, 2.4303750772814773e-06, 8.891804070295328e-06, 1.3318102018055577e-06, 0.0002623315347479568, 2.619584061919865e-06], [2.132061476325401e-05, 0.006726982357699309, 1.6309442212289115e-07, 1.4803897361338603e-07, 0.9911513657004245, 5.096017927428817e-06, 0.0017275950824941354, 1.493535678029813e-05, 0.0003250928179494665, 2.7300918565567074e-05], [0.003201227455745516, 5.209552500064021e-05, 0.0005521672827382476, 0.000588410269065294, 0.0003083719320237835, 0.9922868885518121, 0.00025146533396894677, 0.0009218548831595345, 0.0014541751400888854, 0.0003833436263972401], [0.0001008267341976485, 0.015522143972782069, 0.004686962385200609, 0.15854108294441577, 1.3360785170196824e-05, 0.007650881696104826, 6.705665384658586e-05, 1.5266810616487125e-05, 0.0012301002805328904, 0.8121723177371329], [0.00011946788722919007, 0.008622694182905458, 0.011614769018876774, 0.21017227919200915, 2.5441533841388683e-05, 0.00023606941742385185, 2.89954245196257e-05, 0.00020943835490911105, 0.003610024703878786, 0.7653608202844067], [0.003988518872521902, 0.0070651651588569155, 0.0016796082151826499, 0.009096561829957842, 0.0010811807337679797, 0.03128763426342124, 0.0001991233469207809, 0.2143671058984384, 0.7086974821061695, 0.02253761957476281], [1.553931833179446e-07, 9.907479903280388e-07, 0.00014407075671562099, 0.9997480394038383, 5.562774619016227e-10, 4.854937186086913e-05, 1.7044487276714136e-08, 8.327784381285613e-07, 4.294166699806191e-06, 5.304978050891113e-05], [0.0007154151851907304, 0.006257598497595041, 0.029289299694914155, 0.001501361296402048, 0.003644570442518192, 0.00011685485863155385, 0.000412467426386132, 0.020078448868374184, 0.9376987361270488, 0.0002852476029392177]], "sim": {"pixels": [[0, 0, 5, 13, 9, 1, 0, 0, 0, 0, 13, 15, 10, 15, 5, 0, 0, 3, 15, 2, 0, 11, 8, 0, 0, 4, 12, 0, 0, 8, 8, 0, 0, 5, 8, 0, 0, 9, 8, 0, 0, 4, 11, 0, 1, 12, 7, 0, 0, 2, 14, 5, 10, 12, 0, 0, 0, 0, 6, 13, 10, 0, 0, 0], [0, 0, 0, 4, 15, 12, 0, 0, 0, 0, 3, 16, 15, 14, 0, 0, 0, 0, 8, 13, 8, 16, 0, 0, 0, 0, 1, 6, 15, 11, 0, 0, 0, 1, 8, 13, 15, 1, 0, 0, 0, 9, 16, 16, 5, 0, 0, 0, 0, 3, 13, 16, 16, 11, 5, 0, 0, 0, 0, 3, 11, 16, 9, 0], [0, 0, 0, 1, 11, 0, 0, 0, 0, 0, 0, 7, 8, 0, 0, 0, 0, 0, 1, 13, 6, 2, 2, 0, 0, 0, 7, 15, 0, 9, 8, 0, 0, 5, 16, 10, 0, 16, 6, 0, 0, 4, 15, 16, 13, 16, 1, 0, 0, 0, 0, 3, 15, 10, 0, 0, 0, 0, 0, 2, 16, 4, 0, 0], [0, 0, 12, 10, 0, 0, 0, 0, 0, 0, 14, 16, 16, 14, 0, 0, 0, 0, 13, 16, 15, 10, 1, 0, 0, 0, 11, 16, 16, 7, 0, 0, 0, 0, 0, 4, 7, 16, 7, 0, 0, 0, 0, 0, 4, 16, 9, 0, 0, 0, 5, 4, 12, 16, 4, 0, 0, 0, 9, 16, 16, 10, 0, 0], [0, 0, 7, 8, 13, 16, 15, 1, 0, 0, 7, 7, 4, 11, 12, 0, 0, 0, 0, 0, 8, 13, 1, 0, 0, 4, 8, 8, 15, 15, 6, 0, 0, 2, 11, 15, 15, 4, 0, 0, 0, 0, 0, 16, 5, 0, 0, 0, 0, 0, 9, 15, 1, 0, 0, 0, 0, 0, 13, 5, 0, 0, 0, 0], [0, 0, 9, 14, 8, 1, 0, 0, 0, 0, 12, 14, 14, 12, 0, 0, 0, 0, 9, 10, 0, 15, 4, 0, 0, 0, 3, 16, 12, 14, 2, 0, 0, 0, 4, 16, 16, 2, 0, 0, 0, 3, 16, 8, 10, 13, 2, 0, 0, 1, 15, 1, 3, 16, 8, 0, 0, 0, 11, 16, 15, 11, 1, 0], [0, 0, 1, 9, 15, 11, 0, 0, 0, 0, 11, 16, 8, 14, 6, 0, 0, 2, 16, 10, 0, 9, 9, 0, 0, 1, 16, 4, 0, 8, 8, 0, 0, 4, 16, 4, 0, 8, 8, 0, 0, 1, 16, 5, 1, 11, 3, 0, 0, 0, 12, 12, 10, 10, 0, 0, 0, 0, 1, 10, 13, 3, 0, 0], [0, 0, 0, 0, 14, 13, 1, 0, 0, 0, 0, 5, 16, 16, 2, 0, 0, 0, 0, 14, 16, 12, 0, 0, 0, 1, 10, 16, 16, 12, 0, 0, 0, 3, 12, 14, 16, 9, 0, 0, 0, 0, 0, 5, 16, 15, 0, 0, 0, 0, 0, 4, 16, 14, 0, 0, 0, 0, 0, 1, 13, 16, 1, 0], [0, 0, 5, 12, 1, 0, 0, 0, 0, 0, 15, 14, 7, 0, 0, 0, 0, 0, 13, 1, 12, 0, 0, 0, 0, 2, 10, 0, 14, 0, 0, 0, 0, 0, 2, 0, 16, 1, 0, 0, 0, 0, 0, 6, 15, 0, 0, 0, 0, 0, 9, 16, 15, 9, 8, 2, 0, 0, 3, 11, 8, 13, 12, 4], [0, 0, 0, 8, 15, 1, 0, 0, 0, 0, 0, 12, 14, 0, 0, 0, 0, 0, 3, 16, 7, 0, 0, 0, 0, 0, 6, 16, 2, 0, 0, 0, 0, 0, 7, 16, 16, 13, 5, 0, 0, 0, 15, 16, 9, 9, 14, 0, 0, 0, 3, 14, 9, 2, 16, 2, 0, 0, 0, 7, 15, 16, 11, 0], [0, 0, 1, 8, 15, 10, 0, 0, 0, 3, 13, 15, 14, 14, 0, 0, 0, 5, 10, 0, 10, 12, 0, 0, 0, 0, 3, 5, 15, 10, 2, 0, 0, 0, 16, 16, 16, 16, 12, 0, 0, 1, 8, 12, 14, 8, 3, 0, 0, 0, 0, 10, 13, 0, 0, 0, 0, 0, 0, 11, 9, 0, 0, 0], [0, 0, 10, 7, 13, 9, 0, 0, 0, 0, 9, 10, 12, 15, 2, 0, 0, 0, 4, 11, 10, 11, 0, 0, 0, 0, 1, 16, 10, 1, 0, 0, 0, 0, 12, 13, 4, 0, 0, 0, 0, 0, 12, 1, 12, 0, 0, 0, 0, 1, 10, 2, 14, 0, 0, 0, 0, 0, 11, 14, 5, 0, 0, 0], [0, 0, 3, 13, 11, 7, 0, 0, 0, 0, 11, 16, 16, 16, 2, 0, 0, 4, 16, 9, 1, 14, 2, 0, 0, 4, 16, 0, 0, 16, 2, 0, 0, 0, 16, 1, 0, 12, 8, 0, 0, 0, 15, 9, 0, 13, 6, 0, 0, 0, 9, 14, 9, 14, 1, 0, 0, 0, 2, 12, 13, 4, 0, 0], [0, 0, 0, 2, 16, 16, 2, 0, 0, 0, 0, 4, 16, 16, 2, 0, 0, 1, 4, 12, 16, 12, 0, 0, 0, 7, 16, 16, 16, 12, 0, 0, 0, 0, 3, 10, 16, 14, 0, 0, 0, 0, 0, 8, 16, 12, 0, 0, 0, 0, 0, 6, 16, 16, 2, 0, 0, 0, 0, 2, 12, 15, 4, 0], [0, 1, 8, 12, 15, 14, 4, 0, 0, 3, 11, 8, 8, 12, 12, 0, 0, 0, 0, 0, 2, 13, 7, 0, 0, 0, 0, 2, 15, 12, 1, 0, 0, 0, 0, 0, 13, 5, 0, 0, 0, 0, 0, 0, 9, 13, 0, 0, 0, 0, 7, 8, 14, 15, 0, 0, 0, 0, 14, 15, 11, 2, 0, 0], [0, 0, 1, 13, 14, 3, 0, 0, 0, 0, 8, 16, 13, 2, 0, 0, 0, 2, 16, 16, 3, 0, 0, 0, 0, 3, 16, 12, 1, 0, 0, 0, 0, 5, 16, 14, 5, 0, 0, 0, 0, 3, 16, 16, 16, 16, 6, 0, 0, 1, 14, 16, 16, 16, 12, 0, 0, 0, 3, 12, 15, 14, 7, 0], [0, 0, 0, 8, 14, 14, 2, 0, 0, 0, 0, 6, 10, 15, 11, 0, 0, 0, 0, 0, 0, 14, 10, 0, 0, 2, 8, 11, 12, 16, 8, 0, 0, 8, 16, 16, 16, 16, 7, 0, 0, 0, 0, 0, 11, 15, 1, 0, 0, 0, 0, 9, 16, 7, 0, 0, 0, 0, 0, 12, 13, 1, 0, 0], [0, 0, 10, 11, 4, 0, 0, 0, 0, 0, 10, 15, 13, 13, 1, 0, 0, 0, 8, 11, 0, 14, 4, 0, 0, 0, 0, 13, 15, 13, 0, 0, 0, 1, 11, 16, 16, 0, 0, 0, 0, 1, 15, 3, 9, 10, 0, 0, 0, 0, 14, 6, 15, 10, 0, 0, 0, 0, 8, 14, 7, 1, 0, 0], [0, 0, 10, 14, 11, 3, 0, 0, 0, 4, 16, 13, 6, 14, 1, 0, 0, 4, 16, 2, 0, 11, 7, 0, 0, 8, 16, 0, 0, 10, 5, 0, 0, 8, 16, 0, 0, 14, 4, 0, 0, 8, 16, 0, 1, 16, 1, 0, 0, 4, 16, 1, 11, 15, 0, 0, 0, 0, 11, 16, 12, 3, 0, 0], [0, 2, 13, 16, 16, 16, 11, 0, 0, 5, 16, 10, 5, 4, 1, 0, 0, 6, 16, 7, 3, 0, 0, 0, 0, 9, 16, 16, 16, 6, 0, 0, 0, 3, 8, 4, 11, 15, 0, 0, 0, 0, 0, 1, 12, 15, 0, 0, 0, 0, 4, 13, 16, 6, 0, 0, 0, 2, 16, 15, 8, 0, 0, 0], [0, 6, 13, 5, 8, 8, 1, 0, 0, 8, 16, 16, 16, 16, 6, 0, 0, 6, 16, 9, 6, 4, 0, 0, 0, 6, 16, 16, 15, 5, 0, 0, 0, 0, 4, 5, 15, 12, 0, 0, 0, 0, 0, 3, 16, 9, 0, 0, 0, 1, 8, 13, 15, 3, 0, 0, 0, 4, 16, 15, 3, 0, 0, 0], [0, 0, 0, 5, 14, 2, 0, 0, 0, 0, 1, 13, 11, 0, 0, 0, 0, 0, 5, 16, 2, 0, 0, 0, 0, 0, 6, 15, 5, 0, 0, 0, 0, 1, 15, 16, 15, 11, 1, 0, 0, 2, 13, 14, 1, 12, 9, 0, 0, 0, 4, 16, 7, 13, 9, 0, 0, 0, 0, 5, 16, 15, 3, 0], [0, 3, 15, 8, 8, 6, 0, 0, 0, 4, 16, 16, 16, 13, 2, 0, 0, 3, 16, 9, 2, 0, 0, 0, 0, 2, 16, 16, 15, 3, 0, 0, 0, 0, 7, 6, 12, 9, 0, 0, 0, 0, 0, 1, 14, 10, 0, 0, 0, 0, 5, 14, 15, 2, 0, 0, 0, 1, 15, 14, 1, 0, 0, 0], [0, 0, 6, 14, 10, 2, 0, 0, 0, 0, 15, 15, 13, 15, 3, 0, 0, 2, 16, 10, 0, 13, 9, 0, 0, 1, 16, 5, 0, 12, 5, 0, 0, 0, 16, 3, 0, 13, 6, 0, 0, 1, 15, 5, 6, 13, 1, 0, 0, 0, 16, 11, 14, 10, 0, 0, 0, 0, 7, 16, 11, 1, 0, 0], [0, 0, 13, 10, 1, 0, 0, 0, 0, 5, 16, 14, 7, 0, 0, 0, 0, 4, 16, 8, 14, 0, 0, 0, 0, 2, 14, 16, 16, 6, 0, 0, 0, 0, 1, 4, 9, 13, 1, 0, 0, 0, 0, 0, 0, 13, 6, 0, 0, 0, 5, 8, 5, 9, 14, 0, 0, 0, 13, 13, 15, 16, 13, 0], [0, 1, 9, 16, 13, 7, 0, 0, 0, 7, 14, 4, 10, 12, 0, 0, 0, 6, 15, 9, 16, 11, 0, 0, 0, 0, 9, 11, 7, 14, 0, 0, 0, 0, 0, 0, 0, 15, 2, 0, 0, 0, 0, 0, 0, 11, 6, 0, 0, 3, 13, 8, 5, 14, 5, 0, 0, 0, 9, 14, 13, 10, 1, 0], [0, 0, 11, 10, 12, 4, 0, 0, 0, 0, 12, 13, 9, 16, 1, 0, 0, 0, 7, 13, 11, 16, 0, 0, 0, 0, 1, 16, 14, 4, 0, 0, 0, 0, 10, 16, 13, 0, 0, 0, 0, 0, 14, 7, 12, 7, 0, 0, 0, 4, 14, 4, 12, 13, 0, 0, 0, 1, 11, 14, 12, 4, 0, 0], [0, 0, 0, 9, 15, 1, 0, 0, 0, 0, 4, 16, 12, 0, 0, 0, 0, 0, 15, 14, 2, 11, 3, 0, 0, 4, 16, 9, 4, 16, 10, 0, 0, 9, 16, 11, 13, 16, 2, 0, 0, 0, 9, 16, 16, 14, 0, 0, 0, 0, 0, 8, 16, 6, 0, 0, 0, 0, 0, 9, 16, 2, 0, 0], [0, 0, 0, 0, 12, 5, 0, 0, 0, 0, 0, 2, 16, 12, 0, 0, 0, 0, 1, 12, 16, 11, 0, 0, 0, 2, 12, 16, 16, 10, 0, 0, 0, 6, 11, 5, 15, 6, 0, 0, 0, 0, 0, 1, 16, 9, 0, 0, 0, 0, 0, 2, 16, 11, 0, 0, 0, 0, 0, 3, 16, 8, 0, 0], [0, 0, 0, 9, 15, 12, 0, 0, 0, 0, 4, 7, 7, 14, 0, 0, 0, 0, 0, 0, 0, 13, 3, 0, 0, 4, 9, 8, 10, 13, 1, 0, 0, 4, 16, 15, 16, 16, 6, 0, 0, 0, 0, 0, 14, 3, 0, 0, 0, 0, 0, 9, 12, 0, 0, 0, 0, 0, 0, 11, 7, 0, 0, 0], [0, 0, 9, 16, 16, 16, 5, 0, 0, 1, 14, 10, 8, 16, 8, 0, 0, 0, 0, 0, 7, 16, 3, 0, 0, 3, 8, 11, 15, 16, 11, 0, 0, 8, 16, 16, 15, 11, 3, 0, 0, 0, 2, 16, 7, 0, 0, 0, 0, 0, 8, 16, 1, 0, 0, 0, 0, 0, 13, 10, 0, 0, 0, 0], [0, 1, 15, 4, 0, 0, 0, 0, 0, 2, 16, 16, 16, 14, 2, 0, 0, 6, 16, 11, 8, 8, 3, 0, 0, 5, 16, 11, 5, 0, 0, 0, 0, 0, 11, 14, 14, 1, 0, 0, 0, 0, 0, 5, 16, 7, 0, 0, 0, 0, 6, 16, 16, 4, 0, 0, 0, 0, 14, 14, 4, 0, 0, 0], [0, 0, 0, 1, 11, 9, 0, 0, 0, 0, 0, 7, 16, 13, 0, 0, 0, 0, 4, 14, 16, 9, 0, 0, 0, 10, 16, 11, 16, 8, 0, 0, 0, 0, 0, 3, 16, 6, 0, 0, 0, 0, 0, 3, 16, 8, 0, 0, 0, 0, 0, 5, 16, 10, 0, 0, 0, 0, 0, 2, 14, 6, 0, 0], [0, 0, 2, 15, 13, 3, 0, 0, 0, 0, 10, 15, 11, 15, 0, 0, 0, 3, 16, 6, 0, 10, 0, 0, 0, 4, 16, 8, 0, 3, 8, 0, 0, 8, 14, 3, 0, 4, 8, 0, 0, 3, 15, 1, 0, 3, 7, 0, 0, 0, 14, 11, 6, 14, 5, 0, 0, 0, 4, 12, 15, 6, 0, 0], [0, 0, 1, 15, 13, 1, 0, 0, 0, 0, 7, 16, 14, 8, 0, 0, 0, 8, 12, 9, 2, 13, 2, 0, 0, 7, 9, 1, 0, 6, 6, 0, 0, 5, 9, 0, 0, 3, 9, 0, 0, 0, 15, 2, 0, 8, 12, 0, 0, 0, 9, 15, 13, 16, 6, 0, 0, 0, 0, 13, 14, 8, 0, 0], [0, 0, 0, 5, 14, 12, 2, 0, 0, 0, 7, 15, 8, 14, 4, 0, 0, 0, 6, 2, 3, 13, 1, 0, 0, 0, 0, 1, 13, 4, 0, 0, 0, 0, 1, 11, 9, 0, 0, 0, 0, 8, 16, 13, 0, 0, 0, 0, 0, 5, 14, 16, 11, 2, 0, 0, 0, 0, 0, 6, 12, 13, 3, 0], [0, 0, 0, 3, 15, 10, 1, 0, 0, 0, 0, 11, 10, 16, 4, 0, 0, 0, 0, 12, 1, 15, 6, 0, 0, 0, 0, 3, 4, 15, 4, 0, 0, 0, 0, 6, 15, 6, 0, 0, 0, 4, 15, 16, 9, 0, 0, 0, 0, 0, 13, 16, 15, 9, 3, 0, 0, 0, 0, 4, 9, 14, 7, 0], [0, 0, 3, 12, 16, 16, 6, 0, 0, 0, 10, 11, 7, 16, 11, 0, 0, 0, 0, 0, 2, 14, 10, 0, 0, 5, 11, 8, 9, 16, 3, 0, 0, 9, 16, 16, 16, 16, 9, 0, 0, 1, 4, 9, 16, 6, 0, 0, 0, 0, 0, 11, 14, 0, 0, 0, 0, 0, 4, 16, 5, 0, 0, 0], [0, 0, 0, 1, 14, 14, 3, 0, 0, 0, 0, 10, 11, 13, 8, 0, 0, 0, 0, 7, 0, 13, 8, 0, 0, 0, 0, 0, 7, 15, 1, 0, 0, 4, 8, 12, 15, 4, 0, 0, 0, 6, 16, 16, 6, 0, 0, 0, 0, 0, 2, 12, 12, 4, 2, 0, 0, 0, 0, 1, 13, 16, 5, 0], [0, 0, 2, 14, 15, 5, 0, 0, 0, 0, 10, 16, 16, 15, 1, 0, 0, 3, 16, 10, 10, 16, 4, 0, 0, 5, 16, 0, 0, 14, 6, 0, 0, 5, 16, 6, 0, 12, 7, 0, 0, 1, 15, 13, 4, 13, 6, 0, 0, 0, 11, 16, 16, 15, 0, 0, 0, 0, 2, 11, 13, 4, 0, 0], [0, 0, 0, 1, 9, 11, 0, 0, 0, 0, 0, 13, 16, 16, 0, 0, 0, 0, 0, 12, 7, 14, 0, 0, 0, 0, 0, 0, 14, 7, 0, 0, 0, 0, 5, 12, 12, 0, 0, 0, 0, 7, 16, 16, 6, 0, 0, 0, 0, 4, 9, 13, 16, 11, 4, 0, 0, 0, 0, 0, 9, 13, 3, 0], [0, 1, 10, 15, 11, 1, 0, 0, 0, 3, 8, 8, 11, 12, 0, 0, 0, 0, 0, 5, 14, 15, 1, 0, 0, 0, 0, 11, 15, 2, 0, 0, 0, 0, 0, 4, 15, 2, 0, 0, 0, 0, 0, 0, 12, 10, 0, 0, 0, 0, 3, 4, 10, 16, 1, 0, 0, 0, 13, 16, 15, 10, 0, 0], [0, 0, 10, 15, 14, 4, 0, 0, 0, 0, 4, 6, 13, 16, 2, 0, 0, 0, 0, 3, 16, 9, 0, 0, 0, 0, 0, 1, 16, 6, 0, 0, 0, 0, 0, 0, 10, 12, 0, 0, 0, 0, 0, 0, 1, 16, 4, 0, 0, 1, 9, 5, 6, 16, 7, 0, 0, 0, 14, 12, 15, 11, 2, 0], [0, 0, 6, 13, 16, 6, 0, 0, 0, 3, 16, 14, 15, 16, 1, 0, 0, 0, 5, 0, 8, 16, 2, 0, 0, 0, 0, 0, 8, 16, 3, 0, 0, 3, 15, 16, 16, 16, 9, 0, 0, 5, 13, 14, 16, 11, 3, 0, 0, 0, 0, 12, 15, 1, 0, 0, 0, 0, 4, 16, 7, 0, 0, 0], [0, 0, 14, 16, 14, 6, 0, 0, 0, 0, 7, 10, 16, 16, 3, 0, 0, 0, 0, 5, 16, 16, 1, 0, 0, 0, 0, 2, 16, 8, 0, 0, 0, 0, 0, 0, 12, 13, 1, 0, 0, 0, 0, 0, 4, 16, 7, 0, 0, 0, 5, 9, 14, 16, 7, 0, 0, 0, 13, 16, 16, 10, 1, 0], [0, 3, 16, 16, 14, 7, 1, 0, 0, 1, 9, 9, 15, 16, 4, 0, 0, 0, 0, 7, 16, 12, 1, 0, 0, 0, 0, 9, 16, 2, 0, 0, 0, 0, 0, 3, 15, 7, 0, 0, 0, 0, 0, 0, 9, 15, 0, 0, 0, 1, 10, 10, 16, 16, 3, 0, 0, 2, 13, 16, 12, 5, 0, 0], [0, 0, 0, 6, 16, 4, 0, 0, 0, 0, 1, 13, 15, 1, 0, 0, 0, 1, 11, 16, 5, 0, 0, 0, 0, 8, 16, 10, 0, 10, 6, 0, 0, 12, 16, 8, 9, 16, 12, 0, 0, 2, 15, 16, 16, 16, 7, 0, 0, 0, 0, 4, 16, 11, 0, 0, 0, 0, 0, 7, 16, 3, 0, 0], [0, 0, 0, 9, 10, 0, 0, 0, 0, 0, 7, 16, 7, 0, 0, 0, 0, 0, 13, 13, 1, 0, 0, 0, 0, 0, 15, 7, 0, 0, 0, 0, 0, 4, 16, 15, 12, 7, 0, 0, 0, 2, 16, 12, 4, 11, 10, 0, 0, 0, 8, 14, 5, 9, 14, 0, 0, 0, 0, 6, 12, 14, 9, 0], [0, 0, 0, 10, 11, 0, 0, 0, 0, 0, 9, 16, 6, 0, 0, 0, 0, 0, 15, 13, 0, 0, 0, 0, 0, 0, 14, 10, 0, 0, 0, 0, 0, 1, 15, 12, 8, 2, 0, 0, 0, 0, 12, 16, 16, 16, 10, 1, 0, 0, 7, 16, 12, 12, 16, 4, 0, 0, 0, 9, 15, 12, 5, 0], [0, 0, 5, 14, 0, 0, 0, 0, 0, 0, 12, 9, 0, 0, 0, 0, 0, 0, 15, 3, 0, 0, 0, 0, 0, 1, 16, 0, 0, 0, 0, 0, 0, 1, 16, 2, 7, 4, 0, 0, 0, 3, 16, 16, 16, 16, 9, 0, 0, 0, 15, 15, 4, 10, 16, 0, 0, 0, 4, 14, 16, 12, 7, 0], [0, 0, 0, 9, 9, 0, 0, 0, 0, 0, 3, 16, 9, 0, 0, 0, 0, 3, 14, 10, 0, 2, 0, 0, 0, 10, 16, 5, 7, 15, 1, 0, 0, 2, 11, 15, 16, 13, 1, 0, 0, 0, 0, 7, 16, 3, 0, 0, 0, 0, 0, 6, 15, 0, 0, 0, 0, 0, 0, 4, 16, 5, 0, 0], [0, 0, 6, 12, 13, 6, 0, 0, 0, 6, 16, 9, 12, 16, 2, 0, 0, 7, 16, 9, 15, 13, 0, 0, 0, 0, 11, 15, 16, 4, 0, 0, 0, 0, 0, 12, 10, 0, 0, 0, 0, 0, 3, 16, 4, 0, 0, 0, 0, 0, 1, 16, 2, 0, 0, 0, 0, 0, 6, 11, 0, 0, 0, 0], [0, 1, 13, 15, 12, 12, 5, 0, 0, 4, 16, 8, 8, 6, 0, 0, 0, 7, 13, 0, 0, 0, 0, 0, 0, 8, 15, 13, 15, 7, 0, 0, 0, 1, 6, 5, 8, 12, 0, 0, 0, 0, 0, 0, 12, 11, 0, 0, 0, 0, 2, 13, 14, 1, 0, 0, 0, 3, 14, 10, 1, 0, 0, 0], [0, 0, 1, 13, 10, 0, 0, 0, 0, 7, 16, 16, 16, 7, 0, 0, 0, 8, 16, 13, 10, 15, 0, 0, 0, 8, 16, 2, 2, 15, 3, 0, 0, 5, 15, 2, 0, 12, 7, 0, 0, 1, 15, 6, 2, 16, 3, 0, 0, 0, 11, 15, 13, 16, 0, 0, 0, 0, 1, 15, 14, 8, 0, 0], [0, 1, 12, 13, 4, 0, 0, 0, 0, 4, 16, 16, 16, 3, 0, 0, 0, 4, 16, 16, 16, 10, 0, 0, 0, 0, 6, 16, 14, 16, 0, 0, 0, 0, 0, 0, 0, 16, 4, 0, 0, 0, 0, 0, 0, 13, 7, 0, 0, 1, 2, 3, 7, 14, 10, 0, 0, 2, 12, 16, 14, 12, 3, 0], [0, 0, 13, 13, 8, 2, 0, 0, 0, 5, 16, 16, 16, 12, 0, 0, 0, 1, 15, 12, 0, 0, 0, 0, 0, 0, 12, 13, 7, 1, 0, 0, 0, 0, 8, 16, 16, 12, 0, 0, 0, 0, 0, 4, 9, 16, 3, 0, 0, 0, 1, 5, 14, 15, 1, 0, 0, 0, 10, 16, 16, 6, 0, 0], [0, 0, 0, 0, 9, 13, 0, 0, 0, 0, 0, 2, 16, 16, 1, 0, 0, 0, 0, 5, 9, 15, 0, 0, 0, 0, 0, 0, 5, 14, 0, 0, 0, 0, 0, 3, 15, 7, 0, 0, 0, 7, 16, 16, 11, 0, 0, 0, 0, 0, 11, 14, 16, 7, 3, 0, 0, 0, 0, 0, 9, 15, 9, 0], [0, 3, 5, 14, 13, 6, 0, 0, 0, 9, 16, 12, 10, 12, 0, 0, 0, 6, 16, 3, 12, 11, 0, 0, 0, 1, 13, 10, 16, 6, 0, 0, 0, 0, 10, 16, 10, 0, 0, 0, 0, 1, 15, 16, 10, 0, 0, 0, 0, 0, 16, 12, 16, 0, 0, 0, 0, 0, 3, 15, 16, 5, 0, 0], [0, 0, 0, 0, 11, 15, 4, 0, 0, 0, 0, 3, 16, 16, 12, 0, 0, 0, 0, 8, 14, 16, 12, 0, 0, 0, 0, 5, 10, 16, 6, 0, 0, 1, 7, 11, 16, 13, 0, 0, 0, 9, 16, 16, 14, 1, 0, 0, 0, 3, 8, 14, 16, 9, 0, 0, 0, 0, 0, 1, 11, 16, 12, 0], [0, 0, 10, 12, 10, 0, 0, 0, 0, 3, 16, 16, 16, 4, 0, 0, 0, 7, 15, 3, 8, 13, 0, 0, 0, 8, 12, 0, 0, 14, 1, 0, 0, 8, 12, 0, 0, 7, 8, 0, 0, 5, 13, 0, 0, 4, 8, 0, 0, 0, 14, 8, 0, 10, 8, 0, 0, 0, 7, 12, 13, 12, 4, 0], [0, 0, 0, 1, 14, 13, 1, 0, 0, 0, 0, 1, 16, 16, 3, 0, 0, 5, 11, 15, 16, 16, 0, 0, 0, 4, 15, 16, 16, 15, 0, 0, 0, 0, 0, 8, 16, 7, 0, 0, 0, 0, 0, 10, 16, 3, 0, 0, 0, 0, 0, 8, 16, 6, 0, 0, 0, 0, 0, 2, 13, 15, 2, 0], [0, 0, 3, 14, 16, 14, 0, 0, 0, 0, 13, 13, 13, 16, 2, 0, 0, 0, 1, 0, 9, 15, 0, 0, 0, 0, 9, 12, 15, 16, 10, 0, 0, 4, 16, 16, 16, 11, 3, 0, 0, 0, 4, 9, 14, 2, 0, 0, 0, 0, 2, 15, 9, 0, 0, 0, 0, 0, 4, 13, 1, 0, 0, 0], [0, 2, 13, 15, 10, 4, 0, 0, 0, 0, 5, 4, 13, 15, 2, 0, 0, 0, 0, 0, 11, 16, 4, 0, 0, 0, 0, 0, 16, 12, 0, 0, 0, 0, 0, 0, 13, 11, 0, 0, 0, 0, 0, 0, 8, 13, 0, 0, 0, 1, 6, 8, 14, 12, 0, 0, 0, 2, 12, 14, 11, 1, 0, 0], [0, 1, 13, 15, 2, 0, 0, 0, 0, 6, 15, 15, 9, 0, 0, 0, 0, 9, 8, 10, 13, 0, 0, 0, 0, 5, 3, 12, 12, 0, 0, 0, 0, 0, 3, 16, 6, 0, 0, 0, 0, 5, 15, 15, 1, 0, 0, 0, 0, 6, 16, 15, 12, 12, 11, 0, 0, 1, 11, 13, 16, 16, 12, 0], [0, 0, 0, 1, 16, 5, 0, 0, 0, 0, 0, 5, 16, 11, 0, 0, 0, 0, 0, 12, 16, 11, 0, 0, 0, 7, 12, 16, 16, 7, 0, 0, 0, 4, 8, 12, 16, 4, 0, 0, 0, 0, 0, 9, 16, 2, 0, 0, 0, 0, 0, 10, 16, 2, 0, 0, 0, 0, 0, 3, 13, 5, 0, 0], [0, 0, 2, 7, 15, 13, 1, 0, 0, 0, 14, 12, 9, 14, 8, 0, 0, 0, 2, 0, 0, 12, 8, 0, 0, 0, 0, 0, 0, 13, 6, 0, 0, 5, 16, 16, 16, 16, 5, 0, 0, 2, 5, 7, 13, 14, 2, 0, 0, 0, 0, 1, 15, 5, 0, 0, 0, 0, 0, 11, 9, 0, 0, 0], [0, 0, 0, 9, 16, 4, 0, 0, 0, 1, 9, 16, 13, 2, 0, 0, 0, 14, 16, 14, 8, 0, 0, 0, 1, 15, 15, 5, 16, 9, 0, 0, 0, 5, 16, 16, 16, 8, 0, 0, 0, 0, 2, 13, 16, 1, 0, 0, 0, 0, 0, 11, 13, 0, 0, 0, 0, 0, 0, 11, 13, 0, 0, 0], [0, 0, 13, 16, 15, 4, 0, 0, 0, 0, 9, 8, 13, 16, 3, 0, 0, 0, 0, 0, 13, 16, 7, 0, 0, 0, 0, 1, 16, 12, 0, 0, 0, 0, 0, 0, 15, 10, 0, 0, 0, 0, 0, 0, 8, 15, 0, 0, 0, 0, 3, 6, 15, 16, 7, 0, 0, 0, 15, 16, 16, 11, 1, 0], [0, 0, 0, 1, 12, 8, 1, 0, 0, 0, 0, 4, 16, 16, 1, 0, 0, 0, 1, 13, 16, 11, 0, 0, 0, 1, 11, 16, 16, 12, 0, 0, 0, 2, 12, 8, 16, 10, 0, 0, 0, 0, 0, 0, 15, 8, 0, 0, 0, 0, 0, 4, 16, 4, 0, 0, 0, 0, 0, 3, 13, 4, 0, 0], [0, 4, 14, 16, 16, 12, 1, 0, 0, 2, 12, 7, 14, 16, 6, 0, 0, 0, 0, 5, 16, 10, 0, 0, 0, 0, 0, 4, 16, 7, 0, 0, 0, 0, 0, 4, 16, 6, 0, 0, 0, 0, 0, 1, 15, 11, 0, 0, 0, 1, 8, 10, 16, 10, 0, 0, 0, 5, 16, 16, 15, 1, 0, 0], [0, 0, 9, 13, 14, 5, 0, 0, 0, 4, 16, 10, 13, 16, 0, 0, 0, 0, 13, 15, 14, 16, 1, 0, 0, 0, 0, 3, 7, 16, 3, 0, 0, 0, 0, 0, 4, 16, 0, 0, 0, 0, 0, 0, 1, 16, 3, 0, 0, 1, 15, 5, 8, 16, 2, 0, 0, 0, 7, 15, 16, 9, 0, 0], [0, 0, 3, 12, 16, 10, 0, 0, 0, 2, 14, 12, 12, 12, 0, 0, 0, 5, 10, 0, 10, 11, 0, 0, 0, 0, 0, 1, 14, 9, 2, 0, 0, 0, 8, 16, 16, 16, 10, 0, 0, 0, 6, 16, 13, 7, 0, 0, 0, 0, 0, 16, 5, 0, 0, 0, 0, 0, 5, 13, 0, 0, 0, 0], [0, 0, 0, 11, 16, 8, 0, 0, 0, 0, 6, 16, 13, 3, 0, 0, 0, 0, 8, 16, 8, 0, 0, 0, 0, 0, 13, 16, 2, 0, 0, 0, 0, 0, 15, 16, 5, 0, 0, 0, 0, 2, 16, 16, 16, 5, 0, 0, 0, 1, 10, 16, 16, 14, 0, 0, 0, 0, 0, 12, 16, 15, 0, 0], [0, 0, 0, 3, 14, 1, 0, 0, 0, 0, 0, 13, 12, 1, 0, 0, 0, 0, 7, 16, 5, 3, 0, 0, 0, 3, 15, 11, 5, 16, 2, 0, 0, 5, 16, 11, 11, 16, 6, 0, 0, 0, 6, 12, 16, 13, 3, 0, 0, 0, 0, 1, 15, 7, 0, 0, 0, 0, 0, 2, 16, 7, 0, 0], [0, 2, 15, 16, 16, 13, 2, 0, 0, 1, 10, 8, 14, 16, 8, 0, 0, 0, 0, 0, 16, 15, 1, 0, 0, 0, 0, 0, 16, 8, 0, 0, 0, 0, 0, 0, 14, 14, 0, 0, 0, 0, 0, 0, 11, 16, 1, 0, 0, 2, 14, 13, 16, 16, 3, 0, 0, 2, 15, 16, 14, 5, 0, 0], [0, 0, 1, 15, 13, 0, 0, 0, 0, 0, 1, 16, 16, 5, 0, 0, 0, 0, 7, 16, 16, 0, 0, 0, 0, 0, 13, 16, 13, 0, 0, 0, 0, 7, 16, 16, 13, 0, 0, 0, 0, 1, 11, 16, 13, 0, 0, 0, 0, 0, 2, 16, 16, 0, 0, 0, 0, 0, 1, 14, 16, 3, 0, 0], [0, 0, 1, 12, 5, 0, 0, 0, 0, 0, 9, 16, 14, 3, 0, 0, 0, 2, 16, 14, 11, 13, 0, 0, 0, 2, 16, 10, 0, 14, 4, 0, 0, 4, 16, 0, 0, 12, 4, 0, 0, 4, 16, 3, 0, 11, 10, 0, 0, 0, 13, 12, 8, 14, 6, 0, 0, 0, 3, 10, 16, 12, 1, 0], [0, 4, 13, 16, 16, 12, 3, 0, 0, 3, 7, 4, 13, 16, 6, 0, 0, 0, 0, 8, 15, 5, 0, 0, 0, 0, 0, 12, 8, 0, 0, 0, 0, 0, 0, 7, 12, 0, 0, 0, 0, 0, 0, 4, 12, 0, 0, 0, 0, 1, 7, 12, 11, 0, 0, 0, 0, 3, 15, 12, 2, 0, 0, 0], [0, 0, 0, 8, 14, 4, 0, 0, 0, 0, 7, 16, 7, 0, 0, 0, 0, 0, 14, 10, 0, 0, 0, 0, 0, 1, 16, 6, 0, 0, 0, 0, 0, 3, 16, 16, 10, 0, 0, 0, 0, 2, 16, 12, 14, 6, 0, 0, 0, 0, 12, 15, 11, 10, 0, 0, 0, 0, 0, 10, 13, 8, 0, 0], [0, 0, 9, 15, 5, 0, 0, 0, 0, 0, 13, 14, 13, 7, 0, 0, 0, 0, 6, 14, 10, 13, 1, 0, 0, 0, 0, 9, 12, 15, 5, 0, 0, 0, 0, 0, 0, 10, 10, 0, 0, 0, 0, 0, 0, 7, 14, 0, 0, 0, 3, 8, 9, 15, 15, 0, 0, 0, 5, 12, 12, 9, 1, 0], [0, 0, 0, 5, 11, 1, 0, 0, 0, 0, 0, 14, 14, 2, 0, 0, 0, 0, 5, 16, 5, 0, 0, 0, 0, 0, 8, 15, 2, 0, 0, 0, 0, 0, 10, 13, 0, 0, 0, 0, 0, 0, 14, 16, 16, 8, 0, 0, 0, 0, 6, 16, 9, 15, 6, 0, 0, 0, 0, 6, 14, 16, 8, 0], [0, 0, 0, 0, 6, 10, 0, 0, 0, 0, 0, 0, 14, 15, 0, 0, 0, 0, 0, 10, 16, 16, 0, 0, 0, 0, 7, 16, 16, 15, 0, 0, 0, 3, 15, 7, 15, 10, 0, 0, 0, 0, 0, 0, 13, 11, 0, 0, 0, 0, 0, 0, 15, 9, 0, 0, 0, 0, 0, 0, 8, 15, 0, 0], [0, 0, 2, 11, 16, 4, 0, 0, 0, 0, 12, 9, 11, 15, 1, 0, 0, 0, 2, 0, 4, 16, 0, 0, 0, 0, 0, 2, 8, 15, 1, 0, 0, 4, 16, 16, 16, 15, 7, 0, 0, 3, 6, 4, 16, 3, 0, 0, 0, 0, 0, 6, 11, 0, 0, 0, 0, 0, 0, 12, 7, 0, 0, 0], [0, 0, 15, 16, 16, 12, 2, 0, 0, 2, 16, 15, 12, 12, 3, 0, 0, 4, 16, 8, 0, 0, 0, 0, 0, 8, 16, 12, 4, 0, 0, 0, 0, 0, 12, 16, 16, 0, 0, 0, 0, 0, 0, 3, 16, 8, 0, 0, 0, 2, 6, 9, 16, 8, 0, 0, 0, 1, 15, 16, 15, 3, 0, 0], [0, 0, 0, 10, 15, 0, 0, 0, 0, 0, 11, 15, 3, 0, 0, 0, 0, 7, 15, 4, 0, 0, 0, 0, 0, 12, 11, 1, 3, 8, 2, 0, 0, 4, 12, 15, 15, 16, 9, 0, 0, 0, 0, 8, 16, 8, 2, 0, 0, 0, 0, 10, 12, 0, 0, 0, 0, 0, 0, 12, 9, 0, 0, 0], [0, 0, 3, 11, 16, 13, 0, 0, 0, 3, 15, 15, 13, 16, 0, 0, 0, 6, 8, 2, 9, 14, 0, 0, 0, 0, 4, 7, 15, 14, 5, 0, 0, 4, 16, 16, 16, 14, 6, 0, 0, 1, 8, 13, 12, 0, 0, 0, 0, 0, 1, 16, 4, 0, 0, 0, 0, 0, 3, 15, 1, 0, 0, 0], [0, 0, 2, 16, 13, 1, 0, 0, 0, 0, 6, 15, 16, 7, 0, 0, 0, 0, 3, 6, 16, 8, 0, 0, 0, 0, 0, 3, 16, 6, 0, 0, 0, 0, 1, 13, 13, 0, 0, 0, 0, 6, 14, 16, 4, 0, 0, 0, 0, 9, 16, 16, 13, 10, 3, 0, 0, 0, 4, 11, 15, 16, 10, 0], [0, 0, 4, 9, 13, 5, 0, 0, 0, 1, 16, 16, 12, 11, 0, 0, 0, 0, 11, 8, 5, 16, 0, 0, 0, 0, 7, 10, 13, 10, 0, 0, 0, 0, 4, 16, 13, 0, 0, 0, 0, 0, 13, 15, 12, 0, 0, 0, 0, 0, 11, 9, 14, 5, 0, 0, 0, 0, 0, 10, 16, 8, 0, 0], [0, 0, 0, 8, 15, 8, 0, 0, 0, 0, 3, 16, 12, 16, 4, 0, 0, 0, 2, 10, 1, 16, 4, 0, 0, 0, 0, 0, 8, 14, 0, 0, 0, 0, 0, 9, 15, 3, 0, 0, 0, 3, 16, 14, 4, 0, 0, 0, 0, 4, 15, 14, 7, 1, 0, 0, 0, 0, 0, 9, 12, 14, 4, 0], [0, 2, 15, 16, 16, 14, 2, 0, 0, 3, 16, 14, 9, 10, 1, 0, 0, 7, 16, 6, 0, 0, 0, 0, 0, 3, 14, 15, 6, 0, 0, 0, 0, 0, 1, 13, 12, 0, 0, 0, 0, 0, 0, 8, 16, 0, 0, 0, 0, 0, 9, 16, 11, 0, 0, 0, 0, 3, 16, 13, 1, 0, 0, 0], [0, 0, 9, 12, 15, 15, 2, 0, 0, 1, 15, 14, 11, 16, 7, 0, 0, 0, 2, 0, 2, 16, 4, 0, 0, 0, 2, 4, 10, 15, 2, 0, 0, 0, 13, 16, 16, 16, 9, 0, 0, 0, 13, 16, 10, 1, 0, 0, 0, 0, 6, 16, 1, 0, 0, 0, 0, 0, 11, 9, 0, 0, 0, 0], [0, 0, 0, 12, 14, 1, 0, 0, 0, 0, 9, 16, 10, 5, 0, 0, 0, 0, 8, 13, 5, 14, 0, 0, 0, 0, 2, 14, 16, 16, 4, 0, 0, 0, 0, 0, 4, 10, 10, 0, 0, 0, 0, 0, 0, 4, 16, 0, 0, 0, 2, 6, 4, 9, 16, 0, 0, 0, 1, 11, 16, 15, 7, 0], [0, 0, 6, 13, 2, 0, 0, 0, 0, 4, 16, 16, 16, 11, 0, 0, 0, 0, 12, 11, 1, 6, 1, 0, 0, 0, 12, 14, 10, 2, 0, 0, 0, 0, 1, 8, 12, 12, 0, 0, 0, 0, 0, 0, 9, 14, 0, 0, 0, 0, 4, 9, 16, 5, 0, 0, 0, 0, 9, 14, 4, 0, 0, 0], [0, 0, 0, 12, 13, 0, 0, 0, 0, 0, 5, 16, 10, 0, 0, 0, 0, 6, 16, 13, 11, 12, 2, 0, 0, 12, 16, 10, 15, 16, 9, 0, 0, 4, 14, 16, 16, 12, 4, 0, 0, 0, 0, 14, 13, 0, 0, 0, 0, 0, 0, 14, 12, 0, 0, 0, 0, 0, 0, 11, 13, 0, 0, 0], [0, 0, 3, 6, 14, 5, 0, 0, 0, 0, 7, 15, 15, 13, 0, 0, 0, 0, 3, 14, 13, 12, 0, 0, 0, 0, 0, 16, 13, 1, 0, 0, 0, 0, 2, 16, 8, 0, 0, 0, 0, 0, 10, 9, 14, 0, 0, 0, 0, 0, 12, 4, 15, 2, 0, 0, 0, 0, 4, 15, 14, 0, 0, 0], [0, 0, 5, 14, 15, 4, 0, 0, 0, 0, 8, 16, 16, 14, 0, 0, 0, 0, 5, 16, 16, 9, 0, 0, 0, 0, 0, 15, 16, 1, 0, 0, 0, 0, 1, 16, 13, 0, 0, 0, 0, 0, 11, 15, 14, 5, 0, 0, 0, 0, 12, 12, 8, 15, 1, 0, 0, 0, 5, 16, 16, 16, 2, 0], [0, 0, 12, 16, 7, 0, 0, 0, 0, 2, 16, 5, 12, 3, 0, 0, 0, 0, 14, 6, 3, 16, 2, 0, 0, 0, 2, 14, 16, 12, 0, 0, 0, 0, 0, 0, 10, 10, 0, 0, 0, 0, 0, 0, 10, 8, 0, 0, 0, 0, 8, 2, 13, 7, 0, 0, 0, 0, 11, 16, 16, 3, 0, 0], [0, 0, 1, 11, 15, 8, 0, 0, 0, 0, 12, 14, 10, 16, 5, 0, 0, 0, 16, 7, 13, 16, 4, 0, 0, 0, 9, 15, 13, 16, 4, 0, 0, 0, 0, 0, 1, 15, 4, 0, 0, 0, 0, 0, 0, 16, 4, 0, 0, 8, 16, 14, 9, 16, 4, 0, 0, 0, 2, 10, 15, 15, 2, 0], [0, 0, 7, 13, 15, 5, 0, 0, 0, 0, 8, 16, 16, 12, 0, 0, 0, 0, 7, 16, 15, 3, 0, 0, 0, 0, 6, 16, 5, 0, 0, 0, 0, 0, 5, 16, 2, 0, 0, 0, 0, 0, 8, 16, 6, 0, 0, 0, 0, 0, 12, 12, 13, 0, 0, 0, 0, 0, 5, 13, 10, 0, 0, 0], [0, 0, 2, 12, 9, 0, 0, 0, 0, 0, 11, 15, 12, 5, 0, 0, 0, 0, 15, 5, 0, 14, 0, 0, 0, 2, 15, 1, 0, 9, 7, 0, 0, 4, 10, 0, 0, 7, 8, 0, 0, 0, 12, 0, 0, 8, 10, 0, 0, 2, 15, 5, 10, 16, 1, 0, 0, 0, 5, 14, 12, 4, 0, 0], [0, 2, 15, 16, 6, 0, 0, 0, 0, 5, 16, 15, 14, 0, 0, 0, 0, 5, 13, 10, 14, 0, 0, 0, 0, 0, 0, 12, 12, 0, 0, 0, 0, 0, 1, 16, 7, 0, 0, 0, 0, 0, 10, 15, 2, 0, 0, 0, 0, 3, 16, 10, 8, 6, 1, 0, 0, 2, 15, 16, 16, 16, 7, 0], [0, 3, 16, 16, 12, 12, 6, 0, 0, 0, 4, 4, 5, 14, 8, 0, 0, 0, 0, 0, 11, 11, 0, 0, 0, 0, 0, 4, 16, 3, 0, 0, 0, 0, 0, 0, 12, 11, 0, 0, 0, 0, 0, 0, 9, 14, 0, 0, 0, 0, 3, 7, 15, 4, 0, 0, 0, 3, 16, 14, 4, 0, 0, 0], [0, 4, 16, 16, 16, 10, 3, 0, 0, 12, 16, 9, 8, 12, 3, 0, 0, 10, 16, 2, 0, 0, 0, 0, 0, 3, 16, 12, 0, 0, 0, 0, 0, 0, 7, 16, 9, 0, 0, 0, 0, 0, 0, 12, 16, 1, 0, 0, 0, 3, 10, 15, 15, 1, 0, 0, 0, 4, 16, 13, 3, 0, 0, 0], [0, 0, 2, 15, 15, 3, 0, 0, 0, 0, 8, 14, 16, 11, 0, 0, 0, 0, 0, 0, 11, 14, 0, 0, 0, 0, 0, 0, 11, 14, 3, 0, 0, 0, 4, 12, 16, 16, 7, 0, 0, 0, 11, 16, 12, 1, 0, 0, 0, 0, 1, 14, 6, 0, 0, 0, 0, 0, 4, 12, 1, 0, 0, 0], [0, 0, 5, 15, 5, 0, 0, 0, 0, 0, 12, 11, 13, 8, 0, 0, 0, 0, 11, 9, 10, 16, 0, 0, 0, 0, 3, 15, 16, 16, 1, 0, 0, 0, 0, 0, 0, 14, 6, 0, 0, 0, 0, 0, 0, 13, 8, 0, 0, 0, 6, 3, 0, 14, 6, 0, 0, 0, 2, 13, 16, 15, 3, 0], [0, 0, 12, 9, 12, 1, 0, 0, 0, 0, 14, 16, 16, 8, 0, 0, 0, 3, 16, 9, 3, 15, 2, 0, 0, 4, 16, 1, 0, 16, 5, 0, 0, 5, 12, 0, 0, 16, 5, 0, 0, 3, 14, 1, 4, 16, 4, 0, 0, 0, 15, 12, 14, 14, 0, 0, 0, 0, 7, 12, 12, 2, 0, 0], [0, 0, 0, 0, 7, 13, 5, 0, 0, 0, 0, 0, 15, 16, 8, 0, 0, 0, 1, 13, 16, 16, 3, 0, 0, 0, 12, 16, 16, 16, 0, 0, 0, 8, 16, 3, 16, 13, 0, 0, 0, 2, 3, 0, 16, 12, 0, 0, 0, 0, 0, 0, 16, 13, 0, 0, 0, 0, 0, 0, 8, 12, 0, 0], [0, 0, 13, 14, 8, 1, 0, 0, 0, 1, 16, 16, 16, 6, 0, 0, 0, 0, 3, 5, 16, 8, 0, 0, 0, 0, 0, 9, 16, 2, 0, 0, 0, 0, 1, 16, 9, 0, 0, 0, 0, 1, 16, 13, 0, 0, 0, 0, 0, 5, 16, 11, 4, 4, 0, 0, 0, 1, 13, 14, 12, 12, 0, 0], [0, 4, 15, 16, 16, 16, 4, 0, 0, 4, 16, 15, 9, 7, 1, 0, 0, 0, 15, 14, 1, 0, 0, 0, 0, 0, 5, 16, 9, 0, 0, 0, 0, 0, 0, 14, 16, 0, 0, 0, 0, 0, 0, 9, 16, 5, 0, 0, 0, 3, 10, 13, 16, 4, 0, 0, 0, 5, 16, 16, 12, 0, 0, 0], [0, 0, 0, 6, 14, 1, 0, 0, 0, 0, 2, 16, 10, 0, 0, 0, 0, 0, 11, 15, 2, 0, 0, 0, 0, 0, 13, 13, 8, 5, 0, 0, 0, 2, 16, 14, 12, 16, 3, 0, 0, 1, 16, 11, 0, 5, 12, 0, 0, 0, 11, 15, 5, 12, 12, 0, 0, 0, 0, 9, 15, 14, 6, 0], [0, 0, 3, 15, 14, 1, 0, 0, 0, 0, 13, 16, 16, 6, 0, 0, 0, 0, 4, 4, 16, 8, 0, 0, 0, 0, 0, 1, 16, 10, 1, 0, 0, 0, 7, 16, 16, 16, 8, 0, 0, 0, 12, 15, 16, 6, 0, 0, 0, 0, 0, 14, 11, 0, 0, 0, 0, 0, 4, 12, 1, 0, 0, 0], [0, 0, 6, 13, 10, 4, 0, 0, 0, 4, 16, 15, 13, 13, 0, 0, 0, 4, 16, 14, 16, 16, 1, 0, 0, 2, 10, 16, 16, 16, 5, 0, 0, 0, 0, 0, 3, 16, 5, 0, 0, 0, 0, 0, 0, 16, 8, 0, 0, 0, 0, 2, 8, 16, 8, 0, 0, 0, 7, 16, 14, 9, 1, 0], [0, 0, 2, 12, 4, 0, 0, 0, 0, 1, 12, 16, 16, 3, 0, 0, 0, 7, 16, 6, 4, 13, 0, 0, 0, 8, 16, 6, 0, 13, 5, 0, 0, 1, 16, 5, 0, 7, 9, 0, 0, 0, 16, 8, 0, 8, 12, 0, 0, 0, 13, 14, 14, 16, 10, 0, 0, 0, 4, 14, 15, 7, 0, 0], [0, 0, 0, 0, 5, 14, 3, 0, 0, 0, 0, 0, 9, 16, 8, 0, 0, 0, 0, 9, 16, 16, 5, 0, 0, 1, 13, 15, 12, 16, 1, 0, 0, 4, 12, 3, 10, 15, 0, 0, 0, 0, 0, 0, 11, 12, 0, 0, 0, 0, 0, 0, 8, 12, 0, 0, 0, 0, 0, 0, 5, 13, 4, 0], [0, 0, 6, 13, 10, 3, 0, 0, 0, 5, 15, 11, 16, 11, 0, 0, 0, 2, 6, 0, 16, 12, 0, 0, 0, 0, 0, 0, 16, 12, 0, 0, 0, 0, 0, 7, 16, 6, 0, 0, 0, 0, 5, 15, 10, 0, 0, 0, 0, 0, 11, 16, 8, 4, 0, 0, 0, 0, 8, 14, 13, 12, 4, 0], [0, 1, 11, 12, 14, 6, 0, 0, 0, 1, 6, 4, 8, 16, 7, 0, 0, 0, 0, 0, 3, 16, 5, 0, 0, 0, 0, 2, 14, 9, 0, 0, 0, 0, 0, 4, 16, 7, 0, 0, 0, 0, 0, 1, 14, 13, 0, 0, 0, 0, 0, 1, 12, 14, 0, 0, 0, 5, 12, 15, 9, 1, 0, 0], [0, 0, 0, 3, 16, 3, 0, 0, 0, 0, 0, 10, 16, 11, 0, 0, 0, 0, 4, 16, 16, 8, 0, 0, 0, 2, 14, 12, 16, 5, 0, 0, 0, 10, 16, 14, 16, 16, 11, 0, 0, 5, 12, 13, 16, 8, 3, 0, 0, 0, 0, 2, 15, 3, 0, 0, 0, 0, 0, 4, 12, 0, 0, 0], [0, 0, 10, 16, 14, 1, 0, 0, 0, 0, 14, 16, 16, 7, 0, 0, 0, 0, 1, 5, 16, 7, 0, 0, 0, 0, 2, 10, 16, 11, 6, 0, 0, 7, 16, 16, 15, 12, 7, 0, 0, 11, 10, 15, 10, 0, 0, 0, 0, 0, 6, 16, 2, 0, 0, 0, 0, 0, 13, 9, 0, 0, 0, 0], [0, 0, 2, 13, 13, 1, 0, 0, 0, 0, 8, 16, 14, 4, 0, 0, 0, 0, 5, 16, 10, 8, 4, 0, 0, 0, 1, 16, 16, 10, 2, 0, 0, 2, 15, 13, 12, 0, 0, 0, 0, 5, 12, 3, 15, 1, 0, 0, 0, 0, 14, 3, 13, 4, 0, 0, 0, 0, 3, 15, 13, 1, 0, 0], [0, 0, 5, 13, 4, 0, 0, 0, 0, 0, 15, 14, 14, 5, 0, 0, 0, 0, 12, 7, 8, 16, 1, 0, 0, 0, 4, 13, 16, 16, 1, 0, 0, 0, 0, 0, 0, 15, 2, 0, 0, 0, 0, 0, 0, 14, 6, 0, 0, 0, 2, 7, 8, 16, 4, 0, 0, 0, 6, 12, 13, 10, 0, 0]], "y": [0, 2, 4, 5, 7, 8, 0, 1, 2, 6, 7, 8, 0, 1, 3, 6, 7, 8, 0, 5, 5, 6, 5, 0, 9, 9, 8, 4, 1, 7, 7, 5, 1, 0, 0, 2, 2, 7, 2, 0, 2, 3, 3, 7, 3, 3, 4, 6, 6, 6, 4, 9, 5, 0, 9, 5, 2, 8, 2, 0, 1, 7, 3, 2, 1, 7, 4, 3, 1, 3, 9, 7, 6, 4, 3, 1, 0, 3, 6, 9, 6, 1, 7, 5, 4, 7, 2, 8, 2, 5, 7, 9, 5, 4, 8, 8, 9, 9, 8, 0, 2, 3, 5, 7, 9, 0, 1, 2, 5, 6, 7, 9, 0, 1, 2, 3, 4, 7, 8, 9], "expected_curve": [0.14166666666666666, 0.30833333333333335]}}