nur-dev's picture
Publish reproducibility artifacts (part 2)
fe31b61 verified
Raw History Blame Contribute Delete
110 kB
{ "rows": [ { "samples": 64, "model": "additive_full", "stratum": "unseen", "day": null, "records": 65, "top1": 0.09743589743589744, "top3": 0.36923076923076925, "nll": 2.7052516394680084, "confidence": 0.2678716639535774, "top3_mass": 0.5446903242724688 }, { "samples": 64, "model": "additive_full", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.08333333333333333, "top3": 0.4375, "nll": 2.6498329787463515, "confidence": 0.3139366236646697, "top3_mass": 0.6062274244402662 }, { "samples": 64, "model": "additive_full", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.125, "top3": 0.35416666666666663, "nll": 2.7929448974879656, "confidence": 0.19396763506594805, "top3_mass": 0.4453588031762193 }, { "samples": 64, "model": "additive_full", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.35416666666666663, "nll": 2.6746201539675853, "confidence": 0.2854937346369333, "top3_mass": 0.5671318027034806 }, { "samples": 64, "model": "additive_full", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.09803921568627451, "top3": 0.33333333333333326, "nll": 2.7037051813052995, "confidence": 0.27748766253539503, "top3_mass": 0.5591401525053532 }, { "samples": 64, "model": "additive_full", "stratum": "seen", "day": null, "records": 339, "top1": 0.30186823992133727, "top3": 0.6224188790560472, "nll": 2.189344434424096, "confidence": 0.2838412004888323, "top3_mass": 0.566064286224066 }, { "samples": 64, "model": "additive_full", "stratum": "seen", "day": 9, "records": 85, "top1": 0.3490196078431372, "top3": 0.6509803921568628, "nll": 2.051049710373143, "confidence": 0.30972475129987465, "top3_mass": 0.5912781286604109 }, { "samples": 64, "model": "additive_full", "stratum": "seen", "day": 10, "records": 85, "top1": 0.2352941176470588, "top3": 0.5215686274509804, "nll": 2.4919428671027486, "confidence": 0.2406240820147914, "top3_mass": 0.5258750504253592 }, { "samples": 64, "model": "additive_full", "stratum": "seen", "day": 11, "records": 85, "top1": 0.3294117647058823, "top3": 0.6745098039215686, "nll": 2.1003701481929236, "confidence": 0.287126521566589, "top3_mass": 0.5661468685692526 }, { "samples": 64, "model": "additive_full", "stratum": "seen", "day": 12, "records": 84, "top1": 0.2936507936507936, "top3": 0.6428571428571428, "nll": 2.113118233189516, "confidence": 0.2980566881048511, "top3_mass": 0.5811343926104937 }, { "samples": 64, "model": "additive_full", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "additive_full", "stratum": "supported", "records_across_seeds": 195, "top3": 0.36923076923076925 }, { "samples": 64, "model": "additive_full", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.12307692307692308, "accuracy": 0.2916666666666667 }, { "samples": 64, "model": "additive_full", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.020512820512820513, "accuracy": 0.75 }, { "samples": 64, "model": "additive_full", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.0, "accuracy": null }, { "samples": 64, "model": "additive_diag", "stratum": "unseen", "day": null, "records": 65, "top1": 0.041025641025641026, "top3": 0.22564102564102564, "nll": 14.795769929002507, "confidence": 0.6919640705994234, "top3_mass": 0.9133101428228636 }, { "samples": 64, "model": "additive_diag", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.041666666666666664, "top3": 0.22916666666666666, "nll": 14.248002393844894, "confidence": 0.7031485279108601, "top3_mass": 0.9223906812501139 }, { "samples": 64, "model": "additive_diag", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.0625, "top3": 0.16666666666666666, "nll": 17.996434724255273, "confidence": 0.6640754970720933, "top3_mass": 0.8954982491515744 }, { "samples": 64, "model": "additive_diag", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.041666666666666664, "top3": 0.20833333333333331, "nll": 13.638990954862098, "confidence": 0.681304875436822, "top3_mass": 0.9183208569810671 }, { "samples": 64, "model": "additive_diag", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0196078431372549, "top3": 0.2941176470588235, "nll": 13.387658601045098, "confidence": 0.7177177754262424, "top3_mass": 0.9168119226683555 }, { "samples": 64, "model": "additive_diag", "stratum": "seen", "day": null, "records": 339, "top1": 0.17010816125860373, "top3": 0.4475674615284014, "nll": 10.015167775516005, "confidence": 0.7065046685261072, "top3_mass": 0.9289292644422132 }, { "samples": 64, "model": "additive_diag", "stratum": "seen", "day": 9, "records": 85, "top1": 0.1843137254901961, "top3": 0.4511204481792717, "nll": 9.07373271646867, "confidence": 0.7072057686074378, "top3_mass": 0.9265707354876807 }, { "samples": 64, "model": "additive_diag", "stratum": "seen", "day": 10, "records": 85, "top1": 0.16862745098039214, "top3": 0.392156862745098, "nll": 13.763151722510091, "confidence": 0.7277271153800143, "top3_mass": 0.9449059110960412 }, { "samples": 64, "model": "additive_diag", "stratum": "seen", "day": 11, "records": 85, "top1": 0.1843137254901961, "top3": 0.47099874432531635, "nll": 8.850456574072084, "confidence": 0.6821135871196693, "top3_mass": 0.9142404511341209 }, { "samples": 64, "model": "additive_diag", "stratum": "seen", "day": 12, "records": 84, "top1": 0.14285714285714285, "top3": 0.4763321995464852, "nll": 8.353784402269092, "confidence": 0.7090015784076789, "top3_mass": 0.9300127064273054 }, { "samples": 64, "model": "additive_diag", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "additive_diag", "stratum": "supported", "records_across_seeds": 195, "top3": 0.22564102564102564 }, { "samples": 64, "model": "additive_diag", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.7076923076923077, "accuracy": 0.021739130434782608 }, { "samples": 64, "model": "additive_diag", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.517948717948718, "accuracy": 0.0297029702970297 }, { "samples": 64, "model": "additive_diag", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.3282051282051282, "accuracy": 0.046875 }, { "samples": 64, "model": "factorized_neural", "stratum": "unseen", "day": null, "records": 65, "top1": 0.23589743589743586, "top3": 0.6717948717948717, "nll": 4.429629893875045, "confidence": 0.6135165953585808, "top3_mass": 0.9156864830483303 }, { "samples": 64, "model": "factorized_neural", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.3333333333333333, "top3": 0.75, "nll": 2.62978894501559, "confidence": 0.6377474097758559, "top3_mass": 0.9290275218234602 }, { "samples": 64, "model": "factorized_neural", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.125, "top3": 0.3958333333333333, "nll": 9.789497640639205, "confidence": 0.5621522447279443, "top3_mass": 0.8904230991650923 }, { "samples": 64, "model": "factorized_neural", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.22916666666666666, "top3": 0.75, "nll": 3.1261938221750647, "confidence": 0.6229632451766559, "top3_mass": 0.9216198732834624 }, { "samples": 64, "model": "factorized_neural", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.2549019607843137, "top3": 0.784313725490196, "nll": 2.305779798623542, "confidence": 0.6301630766717916, "top3_mass": 0.9213231464581906 }, { "samples": 64, "model": "factorized_neural", "stratum": "seen", "day": null, "records": 339, "top1": 0.5417895771878072, "top3": 0.775811209439528, "nll": 2.911979581604918, "confidence": 0.6532837292329028, "top3_mass": 0.9268126120284738 }, { "samples": 64, "model": "factorized_neural", "stratum": "seen", "day": 9, "records": 85, "top1": 0.6392156862745098, "top3": 0.8509803921568626, "nll": 1.7004297719310166, "confidence": 0.6851936552144112, "top3_mass": 0.9394369865480234 }, { "samples": 64, "model": "factorized_neural", "stratum": "seen", "day": 10, "records": 85, "top1": 0.37254901960784315, "top3": 0.5686274509803921, "nll": 6.128270478418061, "confidence": 0.5933642366495517, "top3_mass": 0.9086387886403716 }, { "samples": 64, "model": "factorized_neural", "stratum": "seen", "day": 11, "records": 85, "top1": 0.596078431372549, "top3": 0.8352941176470587, "nll": 1.8773649793574834, "confidence": 0.6660819512682848, "top3_mass": 0.9275621238428065 }, { "samples": 64, "model": "factorized_neural", "stratum": "seen", "day": 12, "records": 84, "top1": 0.5595238095238094, "top3": 0.8492063492063491, "nll": 1.9303039718929735, "confidence": 0.6686761612346782, "top3_mass": 0.931669691190482 }, { "samples": 64, "model": "factorized_neural", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "factorized_neural", "stratum": "supported", "records_across_seeds": 195, "top3": 0.6717948717948717 }, { "samples": 64, "model": "factorized_neural", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.6871794871794872, "accuracy": 0.1865671641791045 }, { "samples": 64, "model": "factorized_neural", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.3128205128205128, "accuracy": 0.18032786885245902 }, { "samples": 64, "model": "factorized_neural", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.1076923076923077, "accuracy": 0.14285714285714285 }, { "samples": 64, "model": "joint_neural", "stratum": "unseen", "day": null, "records": 65, "top1": 0.0071876271876271875, "top3": 0.07907203907203907, "nll": 7.578354969750123, "confidence": 0.4605407143336895, "top3_mass": 0.6802111846242919 }, { "samples": 64, "model": "joint_neural", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.005291005291005291, "top3": 0.06828703703703703, "nll": 9.111893686770863, "confidence": 0.5198190196387541, "top3_mass": 0.7386307777989312 }, { "samples": 64, "model": "joint_neural", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.0103505291005291, "top3": 0.09609788359788358, "nll": 4.928531879241698, "confidence": 0.31957205655963095, "top3_mass": 0.5702970321868619 }, { "samples": 64, "model": "joint_neural", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.005291005291005291, "top3": 0.06554232804232804, "nll": 8.621220204043993, "confidence": 0.5222552226923262, "top3_mass": 0.7280353083872384 }, { "samples": 64, "model": "joint_neural", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.00778089013383131, "top3": 0.08593215063803297, "nll": 7.647455336050183, "confidence": 0.4793415087904961, "top3_mass": 0.6836657121535568 }, { "samples": 64, "model": "joint_neural", "stratum": "seen", "day": null, "records": 339, "top1": 0.4750667228543335, "top3": 0.6997331085826659, "nll": 2.762500293635357, "confidence": 0.4982190413250134, "top3_mass": 0.7554462606931474 }, { "samples": 64, "model": "joint_neural", "stratum": "seen", "day": 9, "records": 85, "top1": 0.5647058823529412, "top3": 0.7647058823529411, "nll": 2.3266545418766453, "confidence": 0.545085683844873, "top3_mass": 0.7886435789208275 }, { "samples": 64, "model": "joint_neural", "stratum": "seen", "day": 10, "records": 85, "top1": 0.3103641456582633, "top3": 0.520112044817927, "nll": 3.8074366481933835, "confidence": 0.40782955431617907, "top3_mass": 0.7059474674281697 }, { "samples": 64, "model": "joint_neural", "stratum": "seen", "day": 11, "records": 85, "top1": 0.5411764705882353, "top3": 0.7725490196078431, "nll": 2.3963065146833906, "confidence": 0.5137204837938891, "top3_mass": 0.7591793480596293 }, { "samples": 64, "model": "joint_neural", "stratum": "seen", "day": 12, "records": 84, "top1": 0.484126984126984, "top3": 0.7420634920634921, "nll": 2.5167118409326816, "confidence": 0.5265740314643992, "top3_mass": 0.7681642720267111 }, { "samples": 64, "model": "joint_neural", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "joint_neural", "stratum": "supported", "records_across_seeds": 195, "top3": 0.07907203907203907 }, { "samples": 64, "model": "joint_neural", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.39487179487179486, "accuracy": 0.0 }, { "samples": 64, "model": "joint_neural", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.26666666666666666, "accuracy": 0.0 }, { "samples": 64, "model": "joint_neural", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.14871794871794872, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_nb", "stratum": "unseen", "day": null, "records": 65, "top1": 0.05128205128205128, "top3": 0.29029935581659716, "nll": 42.91473218195748, "confidence": 0.906467353850722, "top3_mass": 0.9977711548307759 }, { "samples": 64, "model": "factorized_nb", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.041666666666666664, "top3": 0.34423747947454847, "nll": 39.99626509140771, "confidence": 0.8968154467193972, "top3_mass": 0.9979913420655012 }, { "samples": 64, "model": "factorized_nb", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.0625, "top3": 0.213439039408867, "nll": 58.1222590329112, "confidence": 0.8962163235731074, "top3_mass": 0.9992409522973286 }, { "samples": 64, "model": "factorized_nb", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.041666666666666664, "top3": 0.23719724958949095, "nll": 39.058661431923205, "confidence": 0.9227196348307279, "top3_mass": 0.9988752574498261 }, { "samples": 64, "model": "factorized_nb", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0588235294117647, "top3": 0.36185163720660674, "nll": 34.977801348668386, "confidence": 0.9099032657838357, "top3_mass": 0.9951414255881132 }, { "samples": 64, "model": "factorized_nb", "stratum": "seen", "day": null, "records": 339, "top1": 0.3126843657817109, "top3": 0.6100818111803769, "nll": 22.194156930407917, "confidence": 0.9123757336666849, "top3_mass": 0.9977169128173773 }, { "samples": 64, "model": "factorized_nb", "stratum": "seen", "day": 9, "records": 85, "top1": 0.3647058823529412, "top3": 0.6696802859074664, "nll": 17.189823366960677, "confidence": 0.9074550331620637, "top3_mass": 0.9977407473992306 }, { "samples": 64, "model": "factorized_nb", "stratum": "seen", "day": 10, "records": 85, "top1": 0.19215686274509802, "top3": 0.41045590650053115, "nll": 39.614162217680416, "confidence": 0.9252652435448174, "top3_mass": 0.998643637399928 }, { "samples": 64, "model": "factorized_nb", "stratum": "seen", "day": 11, "records": 85, "top1": 0.3333333333333333, "top3": 0.6372259248526997, "nll": 17.145857676203747, "confidence": 0.9005945668270137, "top3_mass": 0.9970133701438374 }, { "samples": 64, "model": "factorized_nb", "stratum": "seen", "day": 12, "records": 84, "top1": 0.36111111111111105, "top3": 0.7243089764641488, "nll": 14.739077645672282, "confidence": 0.9162334763407756, "top3_mass": 0.9974669556540026 }, { "samples": 64, "model": "factorized_nb", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "factorized_nb", "stratum": "supported", "records_across_seeds": 195, "top3": 0.2902993558165972 }, { "samples": 64, "model": "factorized_nb", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.9846153846153847, "accuracy": 0.052083333333333336 }, { "samples": 64, "model": "factorized_nb", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.882051282051282, "accuracy": 0.046511627906976744 }, { "samples": 64, "model": "factorized_nb", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.6871794871794872, "accuracy": 0.05970149253731343 }, { "samples": 64, "model": "factorized_lda", "stratum": "unseen", "day": null, "records": 65, "top1": 0.2358974358974359, "top3": 0.5948717948717948, "nll": 4.057728967050472, "confidence": 0.609057438246369, "top3_mass": 0.9063522434424671 }, { "samples": 64, "model": "factorized_lda", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.29166666666666663, "top3": 0.6041666666666666, "nll": 3.323668952015748, "confidence": 0.65075730991364, "top3_mass": 0.9294590404260523 }, { "samples": 64, "model": "factorized_lda", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.16666666666666666, "top3": 0.5208333333333333, "nll": 6.681514735094668, "confidence": 0.5304536465344599, "top3_mass": 0.8694363322324119 }, { "samples": 64, "model": "factorized_lda", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.20833333333333331, "top3": 0.6041666666666666, "nll": 3.3364354155288103, "confidence": 0.6365700501971762, "top3_mass": 0.9144966934555481 }, { "samples": 64, "model": "factorized_lda", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.2745098039215686, "top3": 0.6470588235294117, "nll": 2.9580280715325316, "confidence": 0.6178963164523277, "top3_mass": 0.9116836921138913 }, { "samples": 64, "model": "factorized_lda", "stratum": "seen", "day": null, "records": 339, "top1": 0.5073746312684365, "top3": 0.7826941986234021, "nll": 2.1541319823271787, "confidence": 0.6156095729157138, "top3_mass": 0.9100986736970845 }, { "samples": 64, "model": "factorized_lda", "stratum": "seen", "day": 9, "records": 85, "top1": 0.6039215686274509, "top3": 0.8588235294117645, "nll": 1.407347051037576, "confidence": 0.6385095395089564, "top3_mass": 0.9179138049541038 }, { "samples": 64, "model": "factorized_lda", "stratum": "seen", "day": 10, "records": 85, "top1": 0.3176470588235294, "top3": 0.592156862745098, "nll": 4.339448128156216, "confidence": 0.590129363358726, "top3_mass": 0.9020575026056967 }, { "samples": 64, "model": "factorized_lda", "stratum": "seen", "day": 11, "records": 85, "top1": 0.5725490196078431, "top3": 0.8392156862745097, "nll": 1.4354558951812746, "confidence": 0.61505317302631, "top3_mass": 0.9092340207076758 }, { "samples": 64, "model": "factorized_lda", "stratum": "seen", "day": 12, "records": 84, "top1": 0.5357142857142857, "top3": 0.8412698412698412, "nll": 1.4257071510361057, "confidence": 0.6187835567552102, "top3_mass": 0.9112023509592403 }, { "samples": 64, "model": "factorized_lda", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "factorized_lda", "stratum": "supported", "records_across_seeds": 195, "top3": 0.5948717948717949 }, { "samples": 64, "model": "factorized_lda", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.717948717948718, "accuracy": 0.17857142857142858 }, { "samples": 64, "model": "factorized_lda", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.29743589743589743, "accuracy": 0.1206896551724138 }, { "samples": 64, "model": "factorized_lda", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.07692307692307693, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_kernel", "stratum": "unseen", "day": null, "records": 65, "top1": 0.03076923076923077, "top3": 0.22564102564102564, "nll": 6.021488392317806, "confidence": 0.43642609704021906, "top3_mass": 0.697166444210037 }, { "samples": 64, "model": "factorized_kernel", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.041666666666666664, "top3": 0.2708333333333333, "nll": 6.981712020464046, "confidence": 0.46465075636875747, "top3_mass": 0.7144575314745634 }, { "samples": 64, "model": "factorized_kernel", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.0625, "top3": 0.20833333333333331, "nll": 4.326893731431082, "confidence": 0.36421276794761265, "top3_mass": 0.6397597231587657 }, { "samples": 64, "model": "factorized_kernel", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.020833333333333332, "top3": 0.1875, "nll": 6.51669265554186, "confidence": 0.45743183592393977, "top3_mass": 0.71817466765056 }, { "samples": 64, "model": "factorized_kernel", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0, "top3": 0.2352941176470588, "nll": 6.246586528333271, "confidence": 0.45805709080995777, "top3_mass": 0.7151498892417754 }, { "samples": 64, "model": "factorized_kernel", "stratum": "seen", "day": null, "records": 339, "top1": 0.4464110127826942, "top3": 0.6450344149459193, "nll": 2.4638715185475624, "confidence": 0.5093889701399161, "top3_mass": 0.7685113931493037 }, { "samples": 64, "model": "factorized_kernel", "stratum": "seen", "day": 9, "records": 85, "top1": 0.5333333333333333, "top3": 0.7294117647058822, "nll": 2.2561270365241994, "confidence": 0.5613217412696829, "top3_mass": 0.8030932390650225 }, { "samples": 64, "model": "factorized_kernel", "stratum": "seen", "day": 10, "records": 85, "top1": 0.2784313725490196, "top3": 0.4862745098039215, "nll": 3.0244756629296106, "confidence": 0.4354934453428863, "top3_mass": 0.7195433739681395 }, { "samples": 64, "model": "factorized_kernel", "stratum": "seen", "day": 11, "records": 85, "top1": 0.5019607843137255, "top3": 0.6980392156862744, "nll": 2.2425292486936304, "confidence": 0.5176323105985691, "top3_mass": 0.770024768433547 }, { "samples": 64, "model": "factorized_kernel", "stratum": "seen", "day": 12, "records": 84, "top1": 0.47222222222222215, "top3": 0.6666666666666665, "nll": 2.330788442798707, "confidence": 0.5232717096962479, "top3_mass": 0.7815374387255202 }, { "samples": 64, "model": "factorized_kernel", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "factorized_kernel", "stratum": "supported", "records_across_seeds": 195, "top3": 0.22564102564102564 }, { "samples": 64, "model": "factorized_kernel", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.31794871794871793, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_kernel", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.20512820512820512, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_kernel", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.08717948717948718, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_ridge", "stratum": "unseen", "day": null, "records": 65, "top1": 0.07179487179487179, "top3": 0.2512820512820513, "nll": 3.8936374068228163, "confidence": 0.16161467209769148, "top3_mass": 0.3986772747416128 }, { "samples": 64, "model": "factorized_ridge", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.08333333333333333, "top3": 0.375, "nll": 3.757296078944873, "confidence": 0.17685296555217445, "top3_mass": 0.4278089705034639 }, { "samples": 64, "model": "factorized_ridge", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.0625, "top3": 0.1875, "nll": 3.761080894386952, "confidence": 0.14334166000284088, "top3_mass": 0.3611179240629979 }, { "samples": 64, "model": "factorized_ridge", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.25, "nll": 3.9339779034302538, "confidence": 0.16105217758860563, "top3_mass": 0.3986744969431073 }, { "samples": 64, "model": "factorized_ridge", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0588235294117647, "top3": 0.196078431372549, "nll": 4.108750200899399, "confidence": 0.16500028447364773, "top3_mass": 0.40661179965010164 }, { "samples": 64, "model": "factorized_ridge", "stratum": "seen", "day": null, "records": 339, "top1": 0.30383480825958703, "top3": 0.5457227138643068, "nll": 2.5416625881982657, "confidence": 0.17462620939738682, "top3_mass": 0.416827363526773 }, { "samples": 64, "model": "factorized_ridge", "stratum": "seen", "day": 9, "records": 85, "top1": 0.3529411764705882, "top3": 0.5921568627450979, "nll": 2.4167808083696514, "confidence": 0.18345919742981434, "top3_mass": 0.4293698144865865 }, { "samples": 64, "model": "factorized_ridge", "stratum": "seen", "day": 10, "records": 85, "top1": 0.2235294117647059, "top3": 0.44313725490196076, "nll": 2.7517157811999766, "confidence": 0.16118413846904117, "top3_mass": 0.39381336731682276 }, { "samples": 64, "model": "factorized_ridge", "stratum": "seen", "day": 11, "records": 85, "top1": 0.3372549019607843, "top3": 0.580392156862745, "nll": 2.4466151952188997, "confidence": 0.1737226534834257, "top3_mass": 0.4165671640430074 }, { "samples": 64, "model": "factorized_ridge", "stratum": "seen", "day": 12, "records": 84, "top1": 0.30158730158730157, "top3": 0.5674603174603174, "nll": 2.5516561391927057, "confidence": 0.18020447485976449, "top3_mass": 0.427686867126555 }, { "samples": 64, "model": "factorized_ridge", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "factorized_ridge", "stratum": "supported", "records_across_seeds": 195, "top3": 0.2512820512820513 }, { "samples": 64, "model": "factorized_ridge", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.0, "accuracy": null }, { "samples": 64, "model": "factorized_ridge", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.0, "accuracy": null }, { "samples": 64, "model": "factorized_ridge", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.0, "accuracy": null }, { "samples": 64, "model": "joint_gaussian", "stratum": "unseen", "day": null, "records": 65, "top1": 0.0, "top3": 0.00761904761904762, "nll": 5.197826003203163, "confidence": 0.3723606374759144, "top3_mass": 0.5871707605841656 }, { "samples": 64, "model": "joint_gaussian", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.0, "top3": 0.004166666666666667, "nll": 5.890948796150754, "confidence": 0.44726399817334184, "top3_mass": 0.6611013832597284 }, { "samples": 64, "model": "joint_gaussian", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.0, "top3": 0.007142857142857143, "nll": 4.011263751449254, "confidence": 0.24176444752032095, "top3_mass": 0.4763038968589209 }, { "samples": 64, "model": "joint_gaussian", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.0, "top3": 0.004166666666666667, "nll": 5.604214330938199, "confidence": 0.41917961534023485, "top3_mass": 0.6255005652718985 }, { "samples": 64, "model": "joint_gaussian", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0, "top3": 0.014565826330532213, "nll": 5.279756479505546, "confidence": 0.38071249761129816, "top3_mass": 0.5858591712777643 }, { "samples": 64, "model": "joint_gaussian", "stratum": "seen", "day": null, "records": 339, "top1": 0.4021632251720747, "top3": 0.6602612726506532, "nll": 2.324701055360709, "confidence": 0.4046705309138936, "top3_mass": 0.6473552229616311 }, { "samples": 64, "model": "joint_gaussian", "stratum": "seen", "day": 9, "records": 85, "top1": 0.4666666666666666, "top3": 0.7176470588235293, "nll": 2.1618157990318405, "confidence": 0.44299805414768795, "top3_mass": 0.676329430501364 }, { "samples": 64, "model": "joint_gaussian", "stratum": "seen", "day": 10, "records": 85, "top1": 0.27843137254901956, "top3": 0.5195518207282913, "nll": 2.710196057618621, "confidence": 0.33910861645915413, "top3_mass": 0.6034486108740916 }, { "samples": 64, "model": "joint_gaussian", "stratum": "seen", "day": 11, "records": 85, "top1": 0.4666666666666666, "top3": 0.7176470588235293, "nll": 2.1815517829104705, "confidence": 0.4120192880318859, "top3_mass": 0.6480524842915467 }, { "samples": 64, "model": "joint_gaussian", "stratum": "seen", "day": 12, "records": 84, "top1": 0.3968253968253968, "top3": 0.6865079365079364, "nll": 2.244294623864297, "confidence": 0.4247928987561676, "top3_mass": 0.6617598321702111 }, { "samples": 64, "model": "joint_gaussian", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "joint_gaussian", "stratum": "supported", "records_across_seeds": 195, "top3": 0.007619047619047619 }, { "samples": 64, "model": "joint_gaussian", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.31794871794871793, "accuracy": 0.0 }, { "samples": 64, "model": "joint_gaussian", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.1794871794871795, "accuracy": 0.0 }, { "samples": 64, "model": "joint_gaussian", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.08717948717948718, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_logistic", "stratum": "unseen", "day": null, "records": 65, "top1": 0.08717948717948718, "top3": 0.31794871794871793, "nll": 3.301105921648819, "confidence": 0.27400361906566534, "top3_mass": 0.588542218281554 }, { "samples": 64, "model": "factorized_logistic", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.10416666666666666, "top3": 0.4583333333333333, "nll": 3.298263140192685, "confidence": 0.2824954017366445, "top3_mass": 0.6190831790309694 }, { "samples": 64, "model": "factorized_logistic", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.08333333333333333, "top3": 0.22916666666666666, "nll": 3.0187125972523914, "confidence": 0.2411529184064826, "top3_mass": 0.5148486139319096 }, { "samples": 64, "model": "factorized_logistic", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.29166666666666663, "nll": 3.4491611800644035, "confidence": 0.29396594688447286, "top3_mass": 0.6174606890194885 }, { "samples": 64, "model": "factorized_logistic", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0784313725490196, "top3": 0.2941176470588235, "nll": 3.4302173074718536, "confidence": 0.27814158628392033, "top3_mass": 0.6019390868578307 }, { "samples": 64, "model": "factorized_logistic", "stratum": "seen", "day": null, "records": 339, "top1": 0.38544739429695174, "top3": 0.6774827925270404, "nll": 2.182165054708315, "confidence": 0.3035966808262512, "top3_mass": 0.6150424322806329 }, { "samples": 64, "model": "factorized_logistic", "stratum": "seen", "day": 9, "records": 85, "top1": 0.4862745098039215, "top3": 0.7568627450980392, "nll": 1.944572313843207, "confidence": 0.32109399843753383, "top3_mass": 0.6286204629780792 }, { "samples": 64, "model": "factorized_logistic", "stratum": "seen", "day": 10, "records": 85, "top1": 0.26274509803921564, "top3": 0.5411764705882353, "nll": 2.5402656701642035, "confidence": 0.280167009618971, "top3_mass": 0.5955899176079885 }, { "samples": 64, "model": "factorized_logistic", "stratum": "seen", "day": 11, "records": 85, "top1": 0.4274509803921568, "top3": 0.7137254901960783, "nll": 2.050025153801209, "confidence": 0.3036822726429944, "top3_mass": 0.6147433983329477 }, { "samples": 64, "model": "factorized_logistic", "stratum": "seen", "day": 12, "records": 84, "top1": 0.3650793650793651, "top3": 0.6984126984126983, "nll": 2.193935557528407, "confidence": 0.30951304691240134, "top3_mass": 0.621289444464503 }, { "samples": 64, "model": "factorized_logistic", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "factorized_logistic", "stratum": "supported", "records_across_seeds": 195, "top3": 0.31794871794871793 }, { "samples": 64, "model": "factorized_logistic", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.06153846153846154, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_logistic", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.015384615384615385, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_logistic", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.0, "accuracy": null }, { "samples": 64, "model": "factorized_classifier", "stratum": "unseen", "day": null, "records": 65, "top1": 0.05128205128205128, "top3": 0.36410256410256414, "nll": 4.819105922086376, "confidence": 0.5452796687377128, "top3_mass": 0.8428033475434669 }, { "samples": 64, "model": "factorized_classifier", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.041666666666666664, "top3": 0.375, "nll": 4.8606523929318834, "confidence": 0.5514425838686916, "top3_mass": 0.8578318995890505 }, { "samples": 64, "model": "factorized_classifier", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.08333333333333333, "top3": 0.3125, "nll": 4.9037881563551435, "confidence": 0.5713728799058602, "top3_mass": 0.8449862658843946 }, { "samples": 64, "model": "factorized_classifier", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.041666666666666664, "top3": 0.35416666666666663, "nll": 4.929448997451223, "confidence": 0.5299653099187817, "top3_mass": 0.8403715217167804 }, { "samples": 64, "model": "factorized_classifier", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0392156862745098, "top3": 0.4117647058823529, "nll": 4.596450128694255, "confidence": 0.5293344758151761, "top3_mass": 0.8288930938401023 }, { "samples": 64, "model": "factorized_classifier", "stratum": "seen", "day": null, "records": 339, "top1": 0.48672566371681414, "top3": 0.7236971484759095, "nll": 2.259576138741716, "confidence": 0.5526288192581503, "top3_mass": 0.8417840726549156 }, { "samples": 64, "model": "factorized_classifier", "stratum": "seen", "day": 9, "records": 85, "top1": 0.5999999999999999, "top3": 0.807843137254902, "nll": 1.7954573279197075, "confidence": 0.5716253524234861, "top3_mass": 0.8516330957637923 }, { "samples": 64, "model": "factorized_classifier", "stratum": "seen", "day": 10, "records": 85, "top1": 0.30196078431372547, "top3": 0.5333333333333333, "nll": 3.4299281865664377, "confidence": 0.5520046453530802, "top3_mass": 0.8412753194086539 }, { "samples": 64, "model": "factorized_classifier", "stratum": "seen", "day": 11, "records": 85, "top1": 0.5098039215686274, "top3": 0.776470588235294, "nll": 1.9204013487611018, "confidence": 0.5360379957394399, "top3_mass": 0.8315762847942413 }, { "samples": 64, "model": "factorized_classifier", "stratum": "seen", "day": 12, "records": 84, "top1": 0.5357142857142857, "top3": 0.7777777777777777, "nll": 1.8881479483026862, "confidence": 0.5508260747577671, "top3_mass": 0.8426619182481898 }, { "samples": 64, "model": "factorized_classifier", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 64, "model": "factorized_classifier", "stratum": "supported", "records_across_seeds": 195, "top3": 0.3641025641025641 }, { "samples": 64, "model": "factorized_classifier", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.5743589743589743, "accuracy": 0.03571428571428571 }, { "samples": 64, "model": "factorized_classifier", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.23076923076923078, "accuracy": 0.0 }, { "samples": 64, "model": "factorized_classifier", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.03076923076923077, "accuracy": 0.0 }, { "samples": 128, "model": "additive_full", "stratum": "unseen", "day": null, "records": 65, "top1": 0.12307692307692308, "top3": 0.4615384615384615, "nll": 2.786559313564322, "confidence": 0.3992859131791558, "top3_mass": 0.7043455811342321 }, { "samples": 128, "model": "additive_full", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.10416666666666666, "top3": 0.5625, "nll": 2.7766416573568677, "confidence": 0.44945750644682164, "top3_mass": 0.7540644800921116 }, { "samples": 128, "model": "additive_full", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.14583333333333331, "top3": 0.4375, "nll": 2.9223865295824876, "confidence": 0.31458668891905717, "top3_mass": 0.6277173215655086 }, { "samples": 128, "model": "additive_full", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.10416666666666666, "top3": 0.3958333333333333, "nll": 2.7165645306496597, "confidence": 0.4195958154315265, "top3_mass": 0.7248188657213583 }, { "samples": 128, "model": "additive_full", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.1372549019607843, "top3": 0.45098039215686275, "nll": 2.7339336411915696, "confidence": 0.4126673049345083, "top3_mass": 0.7104030644506719 }, { "samples": 128, "model": "additive_full", "stratum": "seen", "day": null, "records": 339, "top1": 0.3746312684365781, "top3": 0.7059980334316617, "nll": 1.8892189567509576, "confidence": 0.384255624706463, "top3_mass": 0.7062429178122351 }, { "samples": 128, "model": "additive_full", "stratum": "seen", "day": 9, "records": 85, "top1": 0.43529411764705883, "top3": 0.7607843137254902, "nll": 1.7079116620086672, "confidence": 0.41032935945571364, "top3_mass": 0.7260527032579823 }, { "samples": 128, "model": "additive_full", "stratum": "seen", "day": 10, "records": 85, "top1": 0.29803921568627456, "top3": 0.5568627450980391, "nll": 2.273655096425015, "confidence": 0.3498772823771228, "top3_mass": 0.6775196892154636 }, { "samples": 128, "model": "additive_full", "stratum": "seen", "day": 11, "records": 85, "top1": 0.39999999999999997, "top3": 0.7725490196078431, "nll": 1.778413562099199, "confidence": 0.38854849684618464, "top3_mass": 0.7054937405287097 }, { "samples": 128, "model": "additive_full", "stratum": "seen", "day": 12, "records": 84, "top1": 0.365079365079365, "top3": 0.7341269841269842, "nll": 1.7957964177771406, "confidence": 0.3883151189020734, "top3_mass": 0.7160205694424342 }, { "samples": 128, "model": "additive_full", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "additive_full", "stratum": "supported", "records_across_seeds": 195, "top3": 0.46153846153846156 }, { "samples": 128, "model": "additive_full", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.2717948717948718, "accuracy": 0.16981132075471697 }, { "samples": 128, "model": "additive_full", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.12307692307692308, "accuracy": 0.375 }, { "samples": 128, "model": "additive_full", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.035897435897435895, "accuracy": 0.42857142857142855 }, { "samples": 128, "model": "additive_diag", "stratum": "unseen", "day": null, "records": 65, "top1": 0.09743589743589744, "top3": 0.2774725274725275, "nll": 18.666995513635634, "confidence": 0.7759177838492293, "top3_mass": 0.9632808879997385 }, { "samples": 128, "model": "additive_diag", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.10416666666666666, "top3": 0.25148809523809523, "nll": 16.413653621240517, "confidence": 0.7942905874941398, "top3_mass": 0.9616314385766398 }, { "samples": 128, "model": "additive_diag", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.08333333333333333, "top3": 0.20833333333333331, "nll": 28.036218133866175, "confidence": 0.753586047246348, "top3_mass": 0.9597221815089552 }, { "samples": 128, "model": "additive_diag", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.29241071428571425, "nll": 15.589850625751916, "confidence": 0.7952216130457916, "top3_mass": 0.9671585185716636 }, { "samples": 128, "model": "additive_diag", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.1176470588235294, "top3": 0.35294117647058815, "nll": 14.86583237015168, "confidence": 0.7614755285658488, "top3_mass": 0.9645331471451097 }, { "samples": 128, "model": "additive_diag", "stratum": "seen", "day": null, "records": 339, "top1": 0.18977384464110125, "top3": 0.5059905255968729, "nll": 11.73109733245523, "confidence": 0.7608781022497457, "top3_mass": 0.9597249545090075 }, { "samples": 128, "model": "additive_diag", "stratum": "seen", "day": 9, "records": 85, "top1": 0.21568627450980388, "top3": 0.5264512701632377, "nll": 9.985076979146122, "confidence": 0.7571324592030254, "top3_mass": 0.9567159224282744 }, { "samples": 128, "model": "additive_diag", "stratum": "seen", "day": 10, "records": 85, "top1": 0.15294117647058825, "top3": 0.39999999999999997, "nll": 18.002840662297956, "confidence": 0.7979183527642478, "top3_mass": 0.9717485509104878 }, { "samples": 128, "model": "additive_diag", "stratum": "seen", "day": 11, "records": 85, "top1": 0.18823529411764706, "top3": 0.5497005698831257, "nll": 9.938153503803218, "confidence": 0.7290551254484098, "top3_mass": 0.9529962986214127 }, { "samples": 128, "model": "additive_diag", "stratum": "seen", "day": 12, "records": 84, "top1": 0.20238095238095236, "top3": 0.5483081163499882, "nll": 8.965785099479799, "confidence": 0.7593889521706041, "top3_mass": 0.9574118328802219 }, { "samples": 128, "model": "additive_diag", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "additive_diag", "stratum": "supported", "records_across_seeds": 195, "top3": 0.2774725274725275 }, { "samples": 128, "model": "additive_diag", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.8666666666666667, "accuracy": 0.09467455621301775 }, { "samples": 128, "model": "additive_diag", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.6461538461538462, "accuracy": 0.11904761904761904 }, { "samples": 128, "model": "additive_diag", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.4307692307692308, "accuracy": 0.11904761904761904 }, { "samples": 128, "model": "factorized_neural", "stratum": "unseen", "day": null, "records": 65, "top1": 0.31794871794871793, "top3": 0.7282051282051283, "nll": 4.7858047996485595, "confidence": 0.6848218064035226, "top3_mass": 0.9499950626094817 }, { "samples": 128, "model": "factorized_neural", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.47916666666666663, "top3": 0.8333333333333333, "nll": 2.6448228254841553, "confidence": 0.7042986171516052, "top3_mass": 0.9621928436152363 }, { "samples": 128, "model": "factorized_neural", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.16666666666666666, "top3": 0.5208333333333333, "nll": 11.30456105275099, "confidence": 0.6320063069952321, "top3_mass": 0.9365293179610614 }, { "samples": 128, "model": "factorized_neural", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.2708333333333333, "top3": 0.7291666666666666, "nll": 3.1382140066801463, "confidence": 0.7157076264068023, "top3_mass": 0.9483172185710682 }, { "samples": 128, "model": "factorized_neural", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.3529411764705882, "top3": 0.8235294117647058, "nll": 2.216220342265395, "confidence": 0.6871303886688668, "top3_mass": 0.9527675874857915 }, { "samples": 128, "model": "factorized_neural", "stratum": "seen", "day": null, "records": 339, "top1": 0.5968534906588003, "top3": 0.8456243854473942, "nll": 2.578519101938684, "confidence": 0.7385156122596523, "top3_mass": 0.9622960755638232 }, { "samples": 128, "model": "factorized_neural", "stratum": "seen", "day": 9, "records": 85, "top1": 0.7137254901960783, "top3": 0.9137254901960784, "nll": 1.1724433646372425, "confidence": 0.7752087518784624, "top3_mass": 0.9697567410887675 }, { "samples": 128, "model": "factorized_neural", "stratum": "seen", "day": 10, "records": 85, "top1": 0.3803921568627451, "top3": 0.6509803921568628, "nll": 6.204719292900048, "confidence": 0.6644762071030531, "top3_mass": 0.946533334730309 }, { "samples": 128, "model": "factorized_neural", "stratum": "seen", "day": 11, "records": 85, "top1": 0.6666666666666666, "top3": 0.9176470588235294, "nll": 1.4361091710284244, "confidence": 0.7522713994879371, "top3_mass": 0.9665755392963411 }, { "samples": 128, "model": "factorized_neural", "stratum": "seen", "day": 12, "records": 84, "top1": 0.626984126984127, "top3": 0.9007936507936507, "nll": 1.4879746443943844, "confidence": 0.7623869295966507, "top3_mass": 0.9663665753729239 }, { "samples": 128, "model": "factorized_neural", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "factorized_neural", "stratum": "supported", "records_across_seeds": 195, "top3": 0.7282051282051282 }, { "samples": 128, "model": "factorized_neural", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.8307692307692308, "accuracy": 0.345679012345679 }, { "samples": 128, "model": "factorized_neural", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.4717948717948718, "accuracy": 0.2826086956521739 }, { "samples": 128, "model": "factorized_neural", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.18461538461538463, "accuracy": 0.16666666666666666 }, { "samples": 128, "model": "joint_neural", "stratum": "unseen", "day": null, "records": 65, "top1": 0.007081807081807082, "top3": 0.06968864468864469, "nll": 18.437054959514278, "confidence": 0.6247180691136638, "top3_mass": 0.8143641270618712 }, { "samples": 128, "model": "joint_neural", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.002976190476190476, "top3": 0.0536954365079365, "nll": 23.06105667603866, "confidence": 0.6988595432867891, "top3_mass": 0.8611707153093504 }, { "samples": 128, "model": "joint_neural", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.012896825396825396, "top3": 0.12152777777777776, "nll": 10.041144850462839, "confidence": 0.43107548123738904, "top3_mass": 0.704445215887433 }, { "samples": 128, "model": "joint_neural", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.006448412698412698, "top3": 0.04340277777777777, "nll": 21.676251961205224, "confidence": 0.718297746928185, "top3_mass": 0.8710358578673391 }, { "samples": 128, "model": "joint_neural", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.006069094304388421, "top3": 0.060690943043884206, "nll": 18.938430386183544, "confidence": 0.6491147146559015, "top3_mass": 0.8204258608232746 }, { "samples": 128, "model": "joint_neural", "stratum": "seen", "day": null, "records": 339, "top1": 0.5832279814580699, "top3": 0.8049936788874842, "nll": 2.4396589444730514, "confidence": 0.648068051250821, "top3_mass": 0.8701887761740914 }, { "samples": 128, "model": "joint_neural", "stratum": "seen", "day": 9, "records": 85, "top1": 0.7019607843137253, "top3": 0.8862745098039215, "nll": 1.5172525306583875, "confidence": 0.6947924428713661, "top3_mass": 0.8962351249257164 }, { "samples": 128, "model": "joint_neural", "stratum": "seen", "day": 10, "records": 85, "top1": 0.36918767507002803, "top3": 0.6222689075630252, "nll": 4.549346656652398, "confidence": 0.5638677634904052, "top3_mass": 0.8330196009097595 }, { "samples": 128, "model": "joint_neural", "stratum": "seen", "day": 11, "records": 85, "top1": 0.6666666666666666, "top3": 0.8705882352941177, "nll": 1.7741868946303019, "confidence": 0.6618197305905147, "top3_mass": 0.8741011124252351 }, { "samples": 128, "model": "joint_neural", "stratum": "seen", "day": 12, "records": 84, "top1": 0.5952380952380952, "top3": 0.8412698412698412, "nll": 1.911637681087763, "confidence": 0.6720746992033813, "top3_mass": 0.8774851056054348 }, { "samples": 128, "model": "joint_neural", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "joint_neural", "stratum": "supported", "records_across_seeds": 195, "top3": 0.06968864468864469 }, { "samples": 128, "model": "joint_neural", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.6461538461538462, "accuracy": 0.0 }, { "samples": 128, "model": "joint_neural", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.4512820512820513, "accuracy": 0.0 }, { "samples": 128, "model": "joint_neural", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.27692307692307694, "accuracy": 0.0 }, { "samples": 128, "model": "factorized_nb", "stratum": "unseen", "day": null, "records": 65, "top1": 0.09743589743589744, "top3": 0.284331186055324, "nll": 48.13746092912729, "confidence": 0.9065427759403374, "top3_mass": 0.9981097824964309 }, { "samples": 128, "model": "factorized_nb", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.125, "top3": 0.3428520114942529, "nll": 45.738489030930296, "confidence": 0.933956161721246, "top3_mass": 0.997760486858071 }, { "samples": 128, "model": "factorized_nb", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.041666666666666664, "top3": 0.19116892446633826, "nll": 63.03011092291135, "confidence": 0.9050554136295755, "top3_mass": 0.9980967245885937 }, { "samples": 128, "model": "factorized_nb", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.23868534482758622, "nll": 45.19546472036351, "confidence": 0.908694873527863, "top3_mass": 0.9993339675231241 }, { "samples": 128, "model": "factorized_nb", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.1372549019607843, "top3": 0.3598956824108954, "nll": 39.1476426827054, "confidence": 0.8801163090625279, "top3_mass": 0.9972986469853755 }, { "samples": 128, "model": "factorized_nb", "stratum": "seen", "day": null, "records": 339, "top1": 0.2772861356932153, "top3": 0.6474405549016473, "nll": 21.442833017263453, "confidence": 0.9294887562809013, "top3_mass": 0.9988629733857499 }, { "samples": 128, "model": "factorized_nb", "stratum": "seen", "day": 9, "records": 85, "top1": 0.3137254901960784, "top3": 0.7102482372259248, "nll": 15.606794335052992, "confidence": 0.9253803872789539, "top3_mass": 0.9985347339373325 }, { "samples": 128, "model": "factorized_nb", "stratum": "seen", "day": 10, "records": 85, "top1": 0.14901960784313725, "top3": 0.4464309861875784, "nll": 41.07799988263615, "confidence": 0.9321181285361739, "top3_mass": 0.9992559969678647 }, { "samples": 128, "model": "factorized_nb", "stratum": "seen", "day": 11, "records": 85, "top1": 0.2980392156862745, "top3": 0.6857577513764126, "nll": 15.767704451304635, "confidence": 0.9283679901300479, "top3_mass": 0.9987350836737942 }, { "samples": 128, "model": "factorized_nb", "stratum": "seen", "day": 12, "records": 84, "top1": 0.34920634920634913, "top3": 0.7485143482680428, "nll": 13.222119118902842, "confidence": 0.9321194687367346, "top3_mass": 0.9989268302208926 }, { "samples": 128, "model": "factorized_nb", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "factorized_nb", "stratum": "supported", "records_across_seeds": 195, "top3": 0.2843311860553239 }, { "samples": 128, "model": "factorized_nb", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.9897435897435898, "accuracy": 0.09844559585492228 }, { "samples": 128, "model": "factorized_nb", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.8666666666666667, "accuracy": 0.07692307692307693 }, { "samples": 128, "model": "factorized_nb", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.7128205128205128, "accuracy": 0.07194244604316546 }, { "samples": 128, "model": "factorized_lda", "stratum": "unseen", "day": null, "records": 65, "top1": 0.28205128205128205, "top3": 0.6564102564102565, "nll": 4.37381785501415, "confidence": 0.6226594081166319, "top3_mass": 0.9254941614556081 }, { "samples": 128, "model": "factorized_lda", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.3958333333333333, "top3": 0.7083333333333333, "nll": 2.9813366645107, "confidence": 0.6450391110420814, "top3_mass": 0.9378408733687948 }, { "samples": 128, "model": "factorized_lda", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.125, "top3": 0.5625, "nll": 8.834442313864965, "confidence": 0.5695354127026289, "top3_mass": 0.9132466781087576 }, { "samples": 128, "model": "factorized_lda", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.25, "top3": 0.6666666666666666, "nll": 3.1956736940163446, "confidence": 0.6666437977427184, "top3_mass": 0.9358711599917441 }, { "samples": 128, "model": "factorized_lda", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.35294117647058815, "top3": 0.6862745098039215, "nll": 2.5949951657439776, "confidence": 0.6101981402224831, "top3_mass": 0.9156341830062225 }, { "samples": 128, "model": "factorized_lda", "stratum": "seen", "day": null, "records": 339, "top1": 0.5801376597836774, "top3": 0.8505408062930186, "nll": 1.98662723118646, "confidence": 0.6437062979717997, "top3_mass": 0.9282823339096133 }, { "samples": 128, "model": "factorized_lda", "stratum": "seen", "day": 9, "records": 85, "top1": 0.6745098039215686, "top3": 0.9294117647058823, "nll": 0.9734109039953873, "confidence": 0.6733018560167467, "top3_mass": 0.9387235535803247 }, { "samples": 128, "model": "factorized_lda", "stratum": "seen", "day": 10, "records": 85, "top1": 0.3647058823529412, "top3": 0.6509803921568628, "nll": 4.820057628972093, "confidence": 0.589226194798963, "top3_mass": 0.9091790786826228 }, { "samples": 128, "model": "factorized_lda", "stratum": "seen", "day": 11, "records": 85, "top1": 0.6431372549019607, "top3": 0.9098039215686274, "nll": 1.0843282878890155, "confidence": 0.6550976981148339, "top3_mass": 0.92876277078777 }, { "samples": 128, "model": "factorized_lda", "stratum": "seen", "day": 12, "records": 84, "top1": 0.6388888888888888, "top3": 0.9126984126984127, "nll": 1.0577845428500907, "confidence": 0.6573600756350465, "top3_mass": 0.9365613325719991 }, { "samples": 128, "model": "factorized_lda", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "factorized_lda", "stratum": "supported", "records_across_seeds": 195, "top3": 0.6564102564102564 }, { "samples": 128, "model": "factorized_lda", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.7230769230769231, "accuracy": 0.3120567375886525 }, { "samples": 128, "model": "factorized_lda", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.35384615384615387, "accuracy": 0.3188405797101449 }, { "samples": 128, "model": "factorized_lda", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.06666666666666667, "accuracy": 0.15384615384615385 }, { "samples": 128, "model": "factorized_kernel", "stratum": "unseen", "day": null, "records": 65, "top1": 0.03076923076923077, "top3": 0.24615384615384617, "nll": 5.316300051864069, "confidence": 0.4089796567394144, "top3_mass": 0.6577881603905575 }, { "samples": 128, "model": "factorized_kernel", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.020833333333333332, "top3": 0.2708333333333333, "nll": 6.187049752340277, "confidence": 0.4493025956520194, "top3_mass": 0.6918674618903067 }, { "samples": 128, "model": "factorized_kernel", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.041666666666666664, "top3": 0.20833333333333331, "nll": 3.461922371491744, "confidence": 0.30143660896758695, "top3_mass": 0.5669963007745659 }, { "samples": 128, "model": "factorized_kernel", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.041666666666666664, "top3": 0.20833333333333331, "nll": 5.870456106672641, "confidence": 0.46915596501383805, "top3_mass": 0.7021096855030582 }, { "samples": 128, "model": "factorized_kernel", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0196078431372549, "top3": 0.2941176470588235, "nll": 5.720508922534696, "confidence": 0.4156091161133427, "top3_mass": 0.669450309099962 }, { "samples": 128, "model": "factorized_kernel", "stratum": "seen", "day": null, "records": 339, "top1": 0.5722713864306784, "top3": 0.7787610619469026, "nll": 1.7035370484779366, "confidence": 0.49888559422638545, "top3_mass": 0.7573332211820869 }, { "samples": 128, "model": "factorized_kernel", "stratum": "seen", "day": 9, "records": 85, "top1": 0.7254901960784315, "top3": 0.8745098039215686, "nll": 1.2902660282580372, "confidence": 0.5657241634206073, "top3_mass": 0.8019281941633623 }, { "samples": 128, "model": "factorized_kernel", "stratum": "seen", "day": 10, "records": 85, "top1": 0.3333333333333333, "top3": 0.5686274509803921, "nll": 2.591686258672908, "confidence": 0.4063719853839861, "top3_mass": 0.6929345364305184 }, { "samples": 128, "model": "factorized_kernel", "stratum": "seen", "day": 11, "records": 85, "top1": 0.6196078431372549, "top3": 0.8431372549019607, "nll": 1.4628537930179126, "confidence": 0.5130087715445121, "top3_mass": 0.7636207627245573 }, { "samples": 128, "model": "factorized_kernel", "stratum": "seen", "day": 12, "records": 84, "top1": 0.611111111111111, "top3": 0.8293650793650793, "nll": 1.4665540790281857, "confidence": 0.5105749787270324, "top3_mass": 0.771010298198336 }, { "samples": 128, "model": "factorized_kernel", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "factorized_kernel", "stratum": "supported", "records_across_seeds": 195, "top3": 0.24615384615384617 }, { "samples": 128, "model": "factorized_kernel", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.3128205128205128, "accuracy": 0.0 }, { "samples": 128, "model": "factorized_kernel", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.1641025641025641, "accuracy": 0.0 }, { "samples": 128, "model": "factorized_kernel", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.09743589743589744, "accuracy": 0.0 }, { "samples": 128, "model": "factorized_ridge", "stratum": "unseen", "day": null, "records": 65, "top1": 0.1076923076923077, "top3": 0.36410256410256414, "nll": 3.620308522945126, "confidence": 0.17084406487121953, "top3_mass": 0.4103784885167152 }, { "samples": 128, "model": "factorized_ridge", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.125, "top3": 0.47916666666666663, "nll": 3.33964135933288, "confidence": 0.1833077082564009, "top3_mass": 0.43335286071445334 }, { "samples": 128, "model": "factorized_ridge", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.14583333333333331, "top3": 0.35416666666666663, "nll": 3.716761296733412, "confidence": 0.15931156467431082, "top3_mass": 0.3779365525325371 }, { "samples": 128, "model": "factorized_ridge", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.375, "nll": 3.577939539157799, "confidence": 0.16847233985142074, "top3_mass": 0.4108801891075339 }, { "samples": 128, "model": "factorized_ridge", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0784313725490196, "top3": 0.2549019607843137, "nll": 3.8335634628145705, "confidence": 0.17219990659500875, "top3_mass": 0.41881694799494673 }, { "samples": 128, "model": "factorized_ridge", "stratum": "seen", "day": null, "records": 339, "top1": 0.3215339233038348, "top3": 0.5899705014749261, "nll": 2.3604777513523505, "confidence": 0.1764189173031722, "top3_mass": 0.40782343230321827 }, { "samples": 128, "model": "factorized_ridge", "stratum": "seen", "day": 9, "records": 85, "top1": 0.36078431372549025, "top3": 0.6313725490196078, "nll": 2.208778546676431, "confidence": 0.1884269554799133, "top3_mass": 0.42213499987960895 }, { "samples": 128, "model": "factorized_ridge", "stratum": "seen", "day": 10, "records": 85, "top1": 0.23137254901960783, "top3": 0.48235294117647054, "nll": 2.6689342104759657, "confidence": 0.16022324157550283, "top3_mass": 0.38220544811571566 }, { "samples": 128, "model": "factorized_ridge", "stratum": "seen", "day": 11, "records": 85, "top1": 0.37254901960784315, "top3": 0.6352941176470588, "nll": 2.261208907201252, "confidence": 0.17567131515171852, "top3_mass": 0.4074165967155741 }, { "samples": 128, "model": "factorized_ridge", "stratum": "seen", "day": 12, "records": 84, "top1": 0.3214285714285714, "top3": 0.611111111111111, "nll": 2.3023049552188883, "confidence": 0.18141290985915393, "top3_mass": 0.4196761279805311 }, { "samples": 128, "model": "factorized_ridge", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "factorized_ridge", "stratum": "supported", "records_across_seeds": 195, "top3": 0.3641025641025641 }, { "samples": 128, "model": "factorized_ridge", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.0, "accuracy": null }, { "samples": 128, "model": "factorized_ridge", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.0, "accuracy": null }, { "samples": 128, "model": "factorized_ridge", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.0, "accuracy": null }, { "samples": 128, "model": "joint_gaussian", "stratum": "unseen", "day": null, "records": 65, "top1": 0.0007326007326007326, "top3": 0.02808302808302808, "nll": 8.683776289280036, "confidence": 0.47664317064361944, "top3_mass": 0.6904258429615099 }, { "samples": 128, "model": "joint_gaussian", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.0, "top3": 0.01450892857142857, "nll": 10.463109312827804, "confidence": 0.5501762141451358, "top3_mass": 0.7462547130434303 }, { "samples": 128, "model": "joint_gaussian", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.002976190476190476, "top3": 0.04637896825396824, "nll": 5.468643657090163, "confidence": 0.32543171680576677, "top3_mass": 0.6021420340670871 }, { "samples": 128, "model": "joint_gaussian", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.0, "top3": 0.01450892857142857, "nll": 9.804143820197483, "confidence": 0.534624723675783, "top3_mass": 0.7310632699145432 }, { "samples": 128, "model": "joint_gaussian", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.0, "top3": 0.03641456582633053, "nll": 8.98065353890324, "confidence": 0.4951813892828406, "top3_mass": 0.6827246776527751 }, { "samples": 128, "model": "joint_gaussian", "stratum": "seen", "day": null, "records": 339, "top1": 0.4966989745750808, "top3": 0.7548461862621154, "nll": 1.9412243688145847, "confidence": 0.5113309621708919, "top3_mass": 0.7733886222881894 }, { "samples": 128, "model": "joint_gaussian", "stratum": "seen", "day": 9, "records": 85, "top1": 0.596078431372549, "top3": 0.8352941176470587, "nll": 1.611282662679589, "confidence": 0.5552280968559002, "top3_mass": 0.8047972955323657 }, { "samples": 128, "model": "joint_gaussian", "stratum": "seen", "day": 10, "records": 85, "top1": 0.357422969187675, "top3": 0.5752100840336134, "nll": 2.618356821210056, "confidence": 0.4406606310118416, "top3_mass": 0.7250116019798449 }, { "samples": 128, "model": "joint_gaussian", "stratum": "seen", "day": 11, "records": 85, "top1": 0.5568627450980392, "top3": 0.8117647058823528, "nll": 1.7282351983874105, "confidence": 0.526111620463152, "top3_mass": 0.778812249009826 }, { "samples": 128, "model": "joint_gaussian", "stratum": "seen", "day": 12, "records": 84, "top1": 0.4761904761904761, "top3": 0.7976190476190476, "nll": 1.8054251551737437, "confidence": 0.5234662686643619, "top3_mass": 0.7850707797776559 }, { "samples": 128, "model": "joint_gaussian", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "joint_gaussian", "stratum": "supported", "records_across_seeds": 195, "top3": 0.02808302808302808 }, { "samples": 128, "model": "joint_gaussian", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.39487179487179486, "accuracy": 0.0 }, { "samples": 128, "model": "joint_gaussian", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.3076923076923077, "accuracy": 0.0 }, { "samples": 128, "model": "joint_gaussian", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.16923076923076924, "accuracy": 0.0 }, { "samples": 128, "model": "factorized_logistic", "stratum": "unseen", "day": null, "records": 65, "top1": 0.1076923076923077, "top3": 0.4307692307692308, "nll": 3.7817498590116956, "confidence": 0.4645334023468606, "top3_mass": 0.7936232085317145 }, { "samples": 128, "model": "factorized_logistic", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.125, "top3": 0.5833333333333333, "nll": 3.7440981101843676, "confidence": 0.47671897075699377, "top3_mass": 0.8155404893983167 }, { "samples": 128, "model": "factorized_logistic", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.10416666666666666, "top3": 0.3958333333333333, "nll": 3.2766895816392716, "confidence": 0.42915523409441214, "top3_mass": 0.739998888027352 }, { "samples": 128, "model": "factorized_logistic", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.35416666666666663, "nll": 4.004430273633833, "confidence": 0.48397430952104453, "top3_mass": 0.8185530107447356 }, { "samples": 128, "model": "factorized_logistic", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.1176470588235294, "top3": 0.392156862745098, "nll": 4.082956081672977, "confidence": 0.46806440721098386, "top3_mass": 0.8000017849314692 }, { "samples": 128, "model": "factorized_logistic", "stratum": "seen", "day": null, "records": 339, "top1": 0.43461160275319566, "top3": 0.7482792527040314, "nll": 1.8028433744980925, "confidence": 0.4914587870628711, "top3_mass": 0.8162383481935278 }, { "samples": 128, "model": "factorized_logistic", "stratum": "seen", "day": 9, "records": 85, "top1": 0.5294117647058822, "top3": 0.8235294117647058, "nll": 1.4503317429687543, "confidence": 0.5167145841671837, "top3_mass": 0.8326853433917029 }, { "samples": 128, "model": "factorized_logistic", "stratum": "seen", "day": 10, "records": 85, "top1": 0.2901960784313725, "top3": 0.5882352941176471, "nll": 2.4678163469415884, "confidence": 0.45086069155978015, "top3_mass": 0.7924621706356182 }, { "samples": 128, "model": "factorized_logistic", "stratum": "seen", "day": 11, "records": 85, "top1": 0.4666666666666667, "top3": 0.8117647058823529, "nll": 1.579481327012168, "confidence": 0.49522601921199116, "top3_mass": 0.8183147905705261 }, { "samples": 128, "model": "factorized_logistic", "stratum": "seen", "day": 12, "records": 84, "top1": 0.45238095238095233, "top3": 0.7698412698412698, "nll": 1.712683375195713, "confidence": 0.5031716517202631, "top3_mass": 0.8215536207951063 }, { "samples": 128, "model": "factorized_logistic", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "factorized_logistic", "stratum": "supported", "records_across_seeds": 195, "top3": 0.4307692307692308 }, { "samples": 128, "model": "factorized_logistic", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.40512820512820513, "accuracy": 0.11392405063291139 }, { "samples": 128, "model": "factorized_logistic", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.14871794871794872, "accuracy": 0.0 }, { "samples": 128, "model": "factorized_logistic", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.020512820512820513, "accuracy": 0.0 }, { "samples": 128, "model": "factorized_classifier", "stratum": "unseen", "day": null, "records": 65, "top1": 0.12307692307692308, "top3": 0.4512820512820513, "nll": 3.929254435005725, "confidence": 0.5129046884375597, "top3_mass": 0.8397141138738512 }, { "samples": 128, "model": "factorized_classifier", "stratum": "unseen", "day": 9, "records": 16, "top1": 0.10416666666666666, "top3": 0.5208333333333333, "nll": 3.865407385126544, "confidence": 0.5290677572448904, "top3_mass": 0.859553154583891 }, { "samples": 128, "model": "factorized_classifier", "stratum": "unseen", "day": 10, "records": 16, "top1": 0.125, "top3": 0.375, "nll": 4.136536066087745, "confidence": 0.5199122362266275, "top3_mass": 0.8330711460726914 }, { "samples": 128, "model": "factorized_classifier", "stratum": "unseen", "day": 11, "records": 16, "top1": 0.08333333333333333, "top3": 0.41666666666666663, "nll": 3.9905205938832484, "confidence": 0.5131103216833921, "top3_mass": 0.8462324391318675 }, { "samples": 128, "model": "factorized_classifier", "stratum": "unseen", "day": 12, "records": 17, "top1": 0.17647058823529407, "top3": 0.4901960784313726, "nll": 3.736594914930089, "confidence": 0.4909035121154595, "top3_mass": 0.8211593861873603 }, { "samples": 128, "model": "factorized_classifier", "stratum": "seen", "day": null, "records": 339, "top1": 0.6617502458210422, "top3": 0.8426745329400196, "nll": 1.5429868239575844, "confidence": 0.5849239206921237, "top3_mass": 0.8717717671284061 }, { "samples": 128, "model": "factorized_classifier", "stratum": "seen", "day": 9, "records": 85, "top1": 0.807843137254902, "top3": 0.9176470588235294, "nll": 1.1046133442102997, "confidence": 0.6249101609228995, "top3_mass": 0.8877857118457841 }, { "samples": 128, "model": "factorized_classifier", "stratum": "seen", "day": 10, "records": 85, "top1": 0.4117647058823529, "top3": 0.6745098039215686, "nll": 2.6039072740177707, "confidence": 0.556564704730889, "top3_mass": 0.8604317161659223 }, { "samples": 128, "model": "factorized_classifier", "stratum": "seen", "day": 11, "records": 85, "top1": 0.7176470588235293, "top3": 0.8823529411764706, "nll": 1.2577026897055186, "confidence": 0.5761943098766678, "top3_mass": 0.867468027192817 }, { "samples": 128, "model": "factorized_classifier", "stratum": "seen", "day": 12, "records": 84, "top1": 0.7103174603174601, "top3": 0.8968253968253967, "nll": 1.2017089541341197, "confidence": 0.5819920142207278, "top3_mass": 0.8713972067160143 }, { "samples": 128, "model": "factorized_classifier", "stratum": "novel_level", "records_across_seeds": 0, "top3": null }, { "samples": 128, "model": "factorized_classifier", "stratum": "supported", "records_across_seeds": 195, "top3": 0.4512820512820513 }, { "samples": 128, "model": "factorized_classifier", "stratum": "selective_unseen", "threshold": 0.5, "coverage": 0.5230769230769231, "accuracy": 0.1568627450980392 }, { "samples": 128, "model": "factorized_classifier", "stratum": "selective_unseen", "threshold": 0.7, "coverage": 0.11282051282051282, "accuracy": 0.09090909090909091 }, { "samples": 128, "model": "factorized_classifier", "stratum": "selective_unseen", "threshold": 0.9, "coverage": 0.005128205128205128, "accuracy": 0.0 } ], "comparisons": [ { "samples": 64, "candidate": "additive_full", "control": "additive_diag", "gain": 0.14522058823529407, "fixed_day_simultaneous_interval": [ 0.03768344056372547, 0.2637875306372556 ], "day_cluster_simultaneous_interval": [ 0.03921568627450983, 0.2083333333333333 ], "day_gains": [ 0.2083333333333333, 0.1874999999999999, 0.14583333333333337, 0.03921568627450983 ] }, { "samples": 64, "candidate": "additive_full", "control": "factorized_neural", "gain": -0.3002450980392156, "fixed_day_simultaneous_interval": [ -0.46017233455882345, -0.13357843137254893 ], "day_cluster_simultaneous_interval": [ -0.45098039215686264, -0.04166666666666674 ], "day_gains": [ -0.31250000000000006, -0.04166666666666674, -0.3958333333333333, -0.45098039215686264 ] }, { "samples": 64, "candidate": "additive_full", "control": "joint_neural", "gain": 0.2908268168378462, "fixed_day_simultaneous_interval": [ 0.15122954963235294, 0.4490091547035483 ], "day_cluster_simultaneous_interval": [ 0.24740118269530031, 0.3692129629629629 ], "day_gains": [ 0.3692129629629629, 0.258068783068783, 0.28862433862433867, 0.24740118269530031 ] }, { "samples": 64, "candidate": "additive_full", "control": "factorized_nb", "gain": 0.08061031524678836, "fixed_day_simultaneous_interval": [ -0.11089530491614752, 0.2728356448991244 ], "day_cluster_simultaneous_interval": [ -0.02851830387327342, 0.14072762725779958 ], "day_gains": [ 0.09326252052545153, 0.14072762725779958, 0.11696941707717573, -0.02851830387327342 ] }, { "samples": 64, "candidate": "additive_full", "control": "factorized_lda", "gain": -0.22426470588235292, "fixed_day_simultaneous_interval": [ -0.3673406862745098, -0.0854775582107839 ], "day_cluster_simultaneous_interval": [ -0.31372549019607837, -0.16666666666666663 ], "day_gains": [ -0.16666666666666669, -0.16666666666666669, -0.24999999999999994, -0.31372549019607837 ] }, { "samples": 64, "candidate": "additive_full", "control": "factorized_kernel", "gain": 0.14430147058823528, "fixed_day_simultaneous_interval": [ -0.01976122089460783, 0.3167915134803943 ], "day_cluster_simultaneous_interval": [ 0.0980392156862745, 0.1666666666666667 ], "day_gains": [ 0.16666666666666663, 0.14583333333333326, 0.1666666666666667, 0.0980392156862745 ] }, { "samples": 64, "candidate": "additive_full", "control": "factorized_ridge", "gain": 0.11764705882352938, "fixed_day_simultaneous_interval": [ 0.02420343137254899, 0.21614602481617665 ], "day_cluster_simultaneous_interval": [ 0.062499999999999944, 0.16666666666666657 ], "day_gains": [ 0.062499999999999944, 0.16666666666666657, 0.10416666666666669, 0.1372549019607843 ] }, { "samples": 64, "candidate": "additive_full", "control": "joint_gaussian", "gain": 0.362281162464986, "fixed_day_simultaneous_interval": [ 0.22231461998862043, 0.509790796787465 ], "day_cluster_simultaneous_interval": [ 0.3187675070028011, 0.4333333333333333 ], "day_gains": [ 0.4333333333333333, 0.34702380952380946, 0.35000000000000003, 0.3187675070028011 ] }, { "samples": 64, "candidate": "additive_full", "control": "factorized_logistic", "gain": 0.0514705882352941, "fixed_day_simultaneous_interval": [ -0.05177734375000001, 0.1554842601102943 ], "day_cluster_simultaneous_interval": [ -0.020833333333333426, 0.12499999999999994 ], "day_gains": [ -0.020833333333333426, 0.12499999999999994, 0.06250000000000006, 0.039215686274509776 ] }, { "samples": 64, "candidate": "additive_full", "control": "factorized_classifier", "gain": 0.006433823529411797, "fixed_day_simultaneous_interval": [ -0.17769646139705883, 0.1962325750612754 ], "day_cluster_simultaneous_interval": [ -0.0784313725490196, 0.0625 ], "day_gains": [ 0.0625, 0.041666666666666574, 5.551115123125783e-17, -0.0784313725490196 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "additive_full", "gain": 0.3002450980392156, "fixed_day_simultaneous_interval": [ 0.13357843137254893, 0.46017233455882417 ], "day_cluster_simultaneous_interval": [ 0.04166666666666674, 0.45098039215686264 ], "day_gains": [ 0.31250000000000006, 0.04166666666666674, 0.3958333333333333, 0.45098039215686264 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "additive_diag", "gain": 0.44546568627450966, "fixed_day_simultaneous_interval": [ 0.30882161458333335, 0.5879296875000006 ], "day_cluster_simultaneous_interval": [ 0.22916666666666663, 0.5416666666666667 ], "day_gains": [ 0.5208333333333334, 0.22916666666666663, 0.5416666666666667, 0.4901960784313725 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "joint_neural", "gain": 0.5910719148770618, "fixed_day_simultaneous_interval": [ 0.4668121908919818, 0.7175138627499611 ], "day_cluster_simultaneous_interval": [ 0.29973544973544974, 0.698381574852163 ], "day_gains": [ 0.681712962962963, 0.29973544973544974, 0.6844576719576719, 0.698381574852163 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "factorized_nb", "gain": 0.380855413286004, "fixed_day_simultaneous_interval": [ 0.25141999125215814, 0.5084680903274564 ], "day_cluster_simultaneous_interval": [ 0.18239429392446632, 0.512802750410509 ], "day_gains": [ 0.4057625205254516, 0.18239429392446632, 0.512802750410509, 0.4224620882835892 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "factorized_lda", "gain": 0.0759803921568627, "fixed_day_simultaneous_interval": [ -0.04978726256127447, 0.21599264705882348 ], "day_cluster_simultaneous_interval": [ -0.12499999999999994, 0.14583333333333337 ], "day_gains": [ 0.14583333333333337, -0.12499999999999994, 0.14583333333333337, 0.13725490196078427 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "factorized_kernel", "gain": 0.4445465686274509, "fixed_day_simultaneous_interval": [ 0.30882238051470584, 0.5882356770833336 ], "day_cluster_simultaneous_interval": [ 0.1875, 0.5625 ], "day_gains": [ 0.4791666666666667, 0.1875, 0.5625, 0.5490196078431371 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "factorized_ridge", "gain": 0.417892156862745, "fixed_day_simultaneous_interval": [ 0.2650103400735294, 0.5674019607843137 ], "day_cluster_simultaneous_interval": [ 0.20833333333333331, 0.588235294117647 ], "day_gains": [ 0.375, 0.20833333333333331, 0.5, 0.588235294117647 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "joint_gaussian", "gain": 0.6625262605042016, "fixed_day_simultaneous_interval": [ 0.5315407338716737, 0.7929958557641814 ], "day_cluster_simultaneous_interval": [ 0.3886904761904762, 0.7697478991596638 ], "day_gains": [ 0.7458333333333333, 0.3886904761904762, 0.7458333333333333, 0.7697478991596638 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "factorized_logistic", "gain": 0.3517156862745097, "fixed_day_simultaneous_interval": [ 0.21231617647058817, 0.49893018535539446 ], "day_cluster_simultaneous_interval": [ 0.16666666666666669, 0.4901960784313724 ], "day_gains": [ 0.29166666666666663, 0.16666666666666669, 0.45833333333333337, 0.4901960784313724 ] }, { "samples": 64, "candidate": "factorized_neural", "control": "factorized_classifier", "gain": 0.3066789215686274, "fixed_day_simultaneous_interval": [ 0.14246285232843134, 0.4664523973651962 ], "day_cluster_simultaneous_interval": [ 0.08333333333333331, 0.39583333333333337 ], "day_gains": [ 0.37500000000000006, 0.08333333333333331, 0.39583333333333337, 0.37254901960784303 ] }, { "samples": 128, "candidate": "additive_full", "control": "additive_diag", "gain": 0.1854101015406162, "fixed_day_simultaneous_interval": [ 0.030144487482493015, 0.34463663395045546 ], "day_cluster_simultaneous_interval": [ 0.0980392156862745, 0.31101190476190466 ], "day_gains": [ 0.31101190476190466, 0.22916666666666669, 0.10342261904761901, 0.0980392156862745 ] }, { "samples": 128, "candidate": "additive_full", "control": "factorized_neural", "gain": -0.2650122549019607, "fixed_day_simultaneous_interval": [ -0.4399509803921568, -0.10569738051470483 ], "day_cluster_simultaneous_interval": [ -0.37254901960784303, -0.08333333333333326 ], "day_gains": [ -0.27083333333333337, -0.08333333333333326, -0.3333333333333333, -0.37254901960784303 ] }, { "samples": 128, "candidate": "additive_full", "control": "joint_neural", "gain": 0.3918741975957049, "fixed_day_simultaneous_interval": [ 0.23187639736884333, 0.5575254478418984 ], "day_cluster_simultaneous_interval": [ 0.3159722222222222, 0.5088045634920634 ], "day_gains": [ 0.5088045634920634, 0.3159722222222222, 0.3524305555555555, 0.39028944911297847 ] }, { "samples": 128, "candidate": "additive_full", "control": "factorized_nb", "gain": 0.17855294057278082, "fixed_day_simultaneous_interval": [ 0.0031240892774135414, 0.36346698434896646 ], "day_cluster_simultaneous_interval": [ 0.09108470974596727, 0.24633107553366174 ], "day_gains": [ 0.21964798850574702, 0.24633107553366174, 0.1571479885057471, 0.09108470974596727 ] }, { "samples": 128, "candidate": "additive_full", "control": "factorized_lda", "gain": -0.19424019607843146, "fixed_day_simultaneous_interval": [ -0.3379296875, -0.054533930759803596 ], "day_cluster_simultaneous_interval": [ -0.2708333333333333, -0.125 ], "day_gains": [ -0.14583333333333348, -0.125, -0.2708333333333333, -0.23529411764705888 ] }, { "samples": 128, "candidate": "additive_full", "control": "factorized_kernel", "gain": 0.2162990196078431, "fixed_day_simultaneous_interval": [ 0.0372233072916667, 0.4010416666666667 ], "day_cluster_simultaneous_interval": [ 0.15686274509803916, 0.2916666666666666 ], "day_gains": [ 0.2916666666666666, 0.22916666666666669, 0.1875, 0.15686274509803916 ] }, { "samples": 128, "candidate": "additive_full", "control": "factorized_ridge", "gain": 0.09589460784313725, "fixed_day_simultaneous_interval": [ -0.050247012867647045, 0.24448606004902032 ], "day_cluster_simultaneous_interval": [ 0.02083333333333337, 0.196078431372549 ], "day_gains": [ 0.0833333333333332, 0.08333333333333337, 0.02083333333333337, 0.196078431372549 ] }, { "samples": 128, "candidate": "additive_full", "control": "joint_gaussian", "gain": 0.43375058356676, "fixed_day_simultaneous_interval": [ 0.2698694257976628, 0.5984603867588718 ], "day_cluster_simultaneous_interval": [ 0.38132440476190477, 0.5479910714285713 ], "day_gains": [ 0.5479910714285713, 0.39112103174603174, 0.38132440476190477, 0.4145658263305322 ] }, { "samples": 128, "candidate": "additive_full", "control": "factorized_logistic", "gain": 0.030330882352941124, "fixed_day_simultaneous_interval": [ -0.08961416207107836, 0.1551778875612747 ], "day_cluster_simultaneous_interval": [ -0.02083333333333348, 0.05882352941176472 ], "day_gains": [ -0.02083333333333348, 0.041666666666666685, 0.041666666666666685, 0.05882352941176472 ] }, { "samples": 128, "candidate": "additive_full", "control": "factorized_classifier", "gain": 0.011029411764705843, "fixed_day_simultaneous_interval": [ -0.168811274509804, 0.19975643382353078 ], "day_cluster_simultaneous_interval": [ -0.03921568627450989, 0.0625 ], "day_gains": [ 0.04166666666666663, 0.0625, -0.020833333333333315, -0.03921568627450989 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "additive_full", "gain": 0.2650122549019607, "fixed_day_simultaneous_interval": [ 0.1056973805147059, 0.4399509803921568 ], "day_cluster_simultaneous_interval": [ 0.08333333333333326, 0.37254901960784303 ], "day_gains": [ 0.27083333333333337, 0.08333333333333326, 0.3333333333333333, 0.37254901960784303 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "additive_diag", "gain": 0.4504223564425769, "fixed_day_simultaneous_interval": [ 0.2874636456144958, 0.6043749042585786 ], "day_cluster_simultaneous_interval": [ 0.31249999999999994, 0.581845238095238 ], "day_gains": [ 0.581845238095238, 0.31249999999999994, 0.43675595238095233, 0.47058823529411753 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "joint_neural", "gain": 0.6568864524976656, "fixed_day_simultaneous_interval": [ 0.502825503736651, 0.7944206344647236 ], "day_cluster_simultaneous_interval": [ 0.39930555555555547, 0.7796378968253967 ], "day_gains": [ 0.7796378968253967, 0.39930555555555547, 0.6857638888888888, 0.7628384687208215 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "factorized_nb", "gain": 0.4435651954747415, "fixed_day_simultaneous_interval": [ 0.292112954691575, 0.6013496142328166 ], "day_cluster_simultaneous_interval": [ 0.329664408866995, 0.49048132183908044 ], "day_gains": [ 0.4904813218390804, 0.329664408866995, 0.49048132183908044, 0.4636337293538103 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "factorized_lda", "gain": 0.07077205882352922, "fixed_day_simultaneous_interval": [ -0.021293083639705945, 0.1786151960784313 ], "day_cluster_simultaneous_interval": [ -0.04166666666666674, 0.13725490196078416 ], "day_gains": [ 0.12499999999999989, -0.04166666666666674, 0.0625, 0.13725490196078416 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "factorized_kernel", "gain": 0.48131127450980377, "fixed_day_simultaneous_interval": [ 0.31969956341911765, 0.6472133501838246 ], "day_cluster_simultaneous_interval": [ 0.31249999999999994, 0.5625 ], "day_gains": [ 0.5625, 0.31249999999999994, 0.5208333333333333, 0.5294117647058822 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "factorized_ridge", "gain": 0.3609068627450979, "fixed_day_simultaneous_interval": [ 0.17601045496323528, 0.540900926776961 ], "day_cluster_simultaneous_interval": [ 0.16666666666666663, 0.568627450980392 ], "day_gains": [ 0.3541666666666666, 0.16666666666666663, 0.3541666666666667, 0.568627450980392 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "joint_gaussian", "gain": 0.6987628384687207, "fixed_day_simultaneous_interval": [ 0.5531442964555613, 0.8327207683203491 ], "day_cluster_simultaneous_interval": [ 0.474454365079365, 0.8188244047619047 ], "day_gains": [ 0.8188244047619047, 0.474454365079365, 0.714657738095238, 0.7871148459383752 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "factorized_logistic", "gain": 0.2953431372549018, "fixed_day_simultaneous_interval": [ 0.13771426930147046, 0.46201018688725526 ], "day_cluster_simultaneous_interval": [ 0.12499999999999994, 0.43137254901960775 ], "day_gains": [ 0.2499999999999999, 0.12499999999999994, 0.375, 0.43137254901960775 ] }, { "samples": 128, "candidate": "factorized_neural", "control": "factorized_classifier", "gain": 0.2760416666666665, "fixed_day_simultaneous_interval": [ 0.1361824065563726, 0.42156977634804027 ], "day_cluster_simultaneous_interval": [ 0.14583333333333326, 0.33333333333333315 ], "day_gains": [ 0.3125, 0.14583333333333326, 0.3125, 0.33333333333333315 ] } ], "gates": [ { "samples": 64, "candidate": "additive_full", "top3": 0.36979166666666663, "substantial_gain": false, "strongest_available": false, "learning_specific": null, "promotion": false, "strongest_observed_model": "factorized_neural", "strongest_top3": 0.6700367647058822 }, { "samples": 64, "candidate": "factorized_neural", "top3": 0.6700367647058822, "substantial_gain": false, "strongest_available": false, "learning_specific": true, "promotion": false, "strongest_observed_model": "factorized_neural", "strongest_top3": 0.6700367647058822 }, { "samples": 128, "candidate": "additive_full", "top3": 0.461703431372549, "substantial_gain": false, "strongest_available": false, "learning_specific": null, "promotion": false, "strongest_observed_model": "factorized_neural", "strongest_top3": 0.7267156862745097 }, { "samples": 128, "candidate": "factorized_neural", "top3": 0.7267156862745097, "substantial_gain": false, "strongest_available": false, "learning_specific": true, "promotion": false, "strongest_observed_model": "factorized_neural", "strongest_top3": 0.7267156862745097 } ], "all_thirteen_days_audited": true, "complete_source_cohort": false, "valid_source_days": 8, "all_four_target_days_retained": true, "secondary_incomplete_cohort_evaluation": true, "confirmation_days": [ 9, 10, 11, 12 ], "one_physical_platform": true, "draws": 20000, "comparison_family": 40, "original_dynamic_mixture_source_gate_remains_failed": true }
{
"rows": [
{
"samples": 64,
"model": "additive_full",
"stratum": "unseen",
"day": null,
"records": 65,
"top1": 0.09743589743589744,
"top3": 0.36923076923076925,
"nll": 2.7052516394680084,
"confidence": 0.2678716639535774,
"top3_mass": 0.5446903242724688
},
{
"samples": 64,
"model": "additive_full",
"stratum": "unseen",
"day": 9,
"records": 16,
"top1": 0.08333333333333333,
"top3": 0.4375,
"nll": 2.6498329787463515,
"confidence": 0.3139366236646697,
"top3_mass": 0.6062274244402662
},
{
"samples": 64,
"model": "additive_full",
"stratum": "unseen",
"day": 10,
"records": 16,
"top1": 0.125,
"top3": 0.35416666666666663,
"nll": 2.7929448974879656,
"confidence": 0.19396763506594805,
"top3_mass": 0.4453588031762193
},
{
"samples": 64,
"model": "additive_full",
"stratum": "unseen",
"day": 11,
"records": 16,
"top1": 0.08333333333333333,
"top3": 0.35416666666666663,
"nll": 2.6746201539675853,
"confidence": 0.2854937346369333,
"top3_mass": 0.5671318027034806
},
{
"samples": 64,
"model": "additive_full",
"stratum": "unseen",
"day": 12,
"records": 17,
"top1": 0.09803921568627451,
"top3": 0.33333333333333326,
"nll": 2.7037051813052995,
"confidence": 0.27748766253539503,
"top3_mass": 0.5591401525053532
},
{
"samples": 64,
"model": "additive_full",
"stratum": "seen",
"day": null,
"records": 339,
"top1": 0.30186823992133727,
"top3": 0.6224188790560472,
"nll": 2.189344434424096,
"confidence": 0.2838412004888323,
"top3_mass": 0.566064286224066
},
{
"samples": 64,
"model": "additive_full",
"stratum": "seen",
"day": 9,
"records": 85,
"top1": 0.3490196078431372,
"top3": 0.6509803921568628,
"nll": 2.051049710373143,
"confidence": 0.30972475129987465,
"top3_mass": 0.5912781286604109
},
{
"samples": 64,
"model": "additive_full",
"stratum": "seen",
"day": 10,
"records": 85,
"top1": 0.2352941176470588,
"top3": 0.5215686274509804,
"nll": 2.4919428671027486,
"confidence": 0.2406240820147914,
"top3_mass": 0.5258750504253592
},
{
"samples": 64,