File size: 6,709 Bytes
03a6c4c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
#!/usr/bin/env python3
"""
example.py β€” Usage demonstrations for hv_chemical_classifier.

NumPy only.
"""

import numpy as np

from hv_chemical_classifier import (
    Hypothesis, EvidenceChannel, ChemicalClassifier,
    adaptive_tempering, true_voi, correlation_discount,
    make_paleo_scenario,
)


def demo_basic():
    print("=" * 72)
    print("Demo 1 β€” basic classification")
    print("=" * 72)
    print()
    hyp, ch = make_paleo_scenario()
    clf = ChemicalClassifier(hyp, saturation_mode='per_species')
    for c in ch:
        clf.add_channel(c)

    # Observe a strong sauropod-like signal
    clf.observe('femur_morphology', 2)
    clf.observe('tooth_shape', 0)
    clf.observe('bone_density', 2)

    r = clf.classify()
    print(f"  Winner: {r['winner']}")
    print(f"  Posterior: " +
          ", ".join(f"{h.name}={p:.3f}"
                    for h, p in zip(clf.hypotheses, r['posterior'])))
    print()


def demo_ambivalence():
    print("=" * 72)
    print("Demo 2 β€” honest refusal via multi-seed bistability")
    print("=" * 72)
    print()

    hyp, ch = make_paleo_scenario()
    clf = ChemicalClassifier(hyp, saturation_mode='per_species')
    for c in ch:
        clf.add_channel(c)
    # Weak, ambiguous signal
    clf.observe('femur_morphology', 1)
    clf.observe('tooth_shape', 1)

    r = clf.classify_multiseed(n_seeds=20, n_steps=500)
    print(f"  Verdict: {r['verdict']}")
    print(f"  Ambiguous: {r['ambiguous']}")
    print(f"  Winner counts: {r['winner_counts']}")
    print()


def demo_tempering():
    print("=" * 72)
    print("Demo 3 β€” adaptive tempering")
    print("=" * 72)
    print()

    # Sharp vs weak specimen
    for label, obs in [
        ('sharp', {'femur_morphology': 2, 'tooth_shape': 0,
                    'bone_density': 2, 'cortical_thickness': 2}),
        ('weak', {'femur_morphology': 1, 'tooth_shape': 1}),
    ]:
        hyp, ch = make_paleo_scenario()
        clf = ChemicalClassifier(hyp, saturation_mode='per_species')
        for c in ch:
            clf.add_channel(c)
        for k, v in obs.items():
            clf.observe(k, v)
        log_lik = clf.raw_log_likelihoods()
        alpha = adaptive_tempering(log_lik)
        r = clf.classify(apply_tempering=True)
        print(f"  {label:<8s}  Ξ± = {alpha:.3f}  winner = {r['winner']}")
    print()


def demo_discount():
    print("=" * 72)
    print("Demo 4 β€” correlation discount")
    print("=" * 72)
    print()

    for label, group in [('independent', None), ('correlated', 'group')]:
        hyp = [Hypothesis('A', prior=0.5), Hypothesis('B', prior=0.5)]
        channels = [
            EvidenceChannel(f'ch{i}',
                             np.array([[0.6, 0.4], [0.4, 0.6]]),
                             cost=1.0, correlation_group=group)
            for i in range(5)
        ]
        clf = ChemicalClassifier(hyp, saturation_mode='per_species')
        for c in channels:
            clf.add_channel(c)
        for c in channels:
            clf.observe(c.name, 1)

        log_lik = clf.raw_log_likelihoods(
            apply_discount=(label == 'correlated'))
        raw_odds = float(log_lik[1] - log_lik[0])
        print(f"  {label:<12s}  raw log-odds = {raw_odds:+.4f}")
    print()


def demo_voi():
    print("=" * 72)
    print("Demo 5 β€” true VOI (which channel to run next?)")
    print("=" * 72)
    print()

    hyp, ch = make_paleo_scenario()
    clf = ChemicalClassifier(hyp, saturation_mode='per_species')
    for c in ch:
        clf.add_channel(c)

    # Partial evidence
    clf.observe('femur_morphology', 1)

    a = clf.effective_rates()
    clf.reset_concentrations()
    traj, _ = clf.evolve(a, 300, 0.02)
    p = traj[-1] / traj[-1].sum()

    candidates = ['tooth_shape', 'bone_density',
                  'cortical_thickness', 'joint_geometry']
    voi = true_voi(clf, p, candidates)

    print(f"  Current posterior: " +
          ", ".join(f"{h.name}={p[i]:.3f}"
                    for i, h in enumerate(clf.hypotheses)))
    print()
    print(f"  {'channel':<22s}  {'VOI/cost':>10s}")
    print("  " + "-" * 36)
    for name, r in sorted(voi.items(), key=lambda x: -x[1]['voi_per_cost']):
        print(f"  {name:<22s}  {r['voi_per_cost']:>10.5f}")
    print()


def demo_messenger():
    print("=" * 72)
    print("Demo 6 β€” pairwise messenger as confidence signal")
    print("=" * 72)
    print()

    # Scenario: clear winner
    hyp, ch = make_paleo_scenario()
    clf = ChemicalClassifier(hyp, saturation_mode='per_species')
    for c in ch:
        clf.add_channel(c)
    clf.observe('femur_morphology', 2)
    clf.observe('tooth_shape', 0)
    clf.observe('bone_density', 2)

    a = clf.effective_rates()
    clf.reset_concentrations()
    traj, M = clf.evolve(a, 500, 0.02, track_messenger=True)

    print(f"  Scenario: clear winner")
    print(f"  Peak messenger: {M.max():.4f} at step {int(np.argmax(M))}")
    print(f"  Final messenger: {M[-1]:.6f}")
    print()

    # Peak occurs during the transition, then decays
    peak_idx = int(np.argmax(M))
    print(f"  Messenger trajectory (every 50 steps):")
    for t in range(0, 500, 50):
        bar = "β–ˆ" * int(M[t] * 40)
        print(f"    step {t:>3d}  M = {M[t]:.4f}  {bar}")
    print()


def demo_stopping():
    print("=" * 72)
    print("Demo 7 β€” entropy stopping vs max-posterior stopping")
    print("=" * 72)
    print()

    hyp, ch = make_paleo_scenario()
    clf = ChemicalClassifier(hyp, saturation_mode='per_species')
    for c in ch:
        clf.add_channel(c)
    clf.observe('femur_morphology', 2)
    clf.observe('tooth_shape', 0)
    clf.observe('bone_density', 2)
    clf.observe('cortical_thickness', 2)

    a = clf.effective_rates()
    clf.reset_concentrations()
    traj, _ = clf.evolve(a, 500, 0.02)

    threshold = 0.85
    stop_step = None
    for t in range(len(traj)):
        H = traj[t]
        p = H / H.sum()
        if p.max() > threshold:
            stop_step = t
            break

    print(f"  Threshold: max posterior > {threshold}")
    print(f"  Stopping step: {stop_step}")
    print()
    print(f"  {'step':>5s}  {'max p':>8s}  {'entropy':>10s}  {'stop?':>6s}")
    print("  " + "-" * 36)
    for t in [0, 50, 100, 150, 200, 300, 499]:
        H = traj[t]
        p = H / H.sum()
        p = np.maximum(p, 1e-12)
        ent = -np.sum(p * np.log(p))
        stopped = "βœ“" if (stop_step is not None and t >= stop_step) else ""
        print(f"  {t:>5d}  {p.max():>8.4f}  {ent:>10.4f}  {stopped:>6s}")
    print()


def main():
    demo_basic()
    demo_ambivalence()
    demo_tempering()
    demo_discount()
    demo_voi()
    demo_messenger()
    demo_stopping()


if __name__ == "__main__":
    main()