{"id":"W4365441057","doi":"10.48550/arxiv.2304.04858","title":"Simulated Annealing in Early Layers Leads to Better Generalization","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Domain Adaptation and Few-Shot Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Compute Canada","keywords":"Initialization; Computer science; Margin (machine learning); Artificial intelligence; Forgetting; Benchmark (surveying); Gradient descent; Transfer of learning; Machine learning; Generalization; Artificial neural network; Mathematics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0003142586,0.0002273943,0.0002364621,0.0006891291,0.0001050669,0.0001954334,0.001010101,0.0002245076,0.00001465707],"category_scores_gemma":[0.00005601966,0.000295244,0.00009967165,0.00146842,0.00002671823,0.000328673,0.0009552693,0.0004458731,0.0003576148],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001993657,"about_ca_system_score_gemma":0.00008355478,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004341195,"about_ca_topic_score_gemma":0.0001210051,"domain_scores_codex":[0.9981962,0.0001753414,0.0002283493,0.0009222528,0.0001172601,0.0003606369],"domain_scores_gemma":[0.9988756,0.00008360405,0.0001403803,0.0006298909,0.0001139655,0.0001565421],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001082892,0.00001316832,0.02554566,0.00001552565,0.00001504224,0.0001957296,0.001074913,0.9620866,0.00007978097,0.01047636,0.000177773,0.0003086051],"study_design_scores_gemma":[0.0003667684,0.00002706924,0.02165274,0.00008304477,0.000009721013,4.482945e-7,0.00004907184,0.9728829,0.00004574264,0.003977588,0.0005465067,0.0003583681],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5239401,0.000004032569,0.4747277,0.0002407842,0.0003366081,0.0001549512,0.000002077849,0.0002627985,0.0003309111],"genre_scores_gemma":[0.99405,0.00001707142,0.002908461,0.0005522016,0.00005363461,6.927303e-7,0.00002442569,0.00002698101,0.00236652],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4718193,"threshold_uncertainty_score":0.99995,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1059978539784337,"score_gpt":0.2160249857094276,"score_spread":0.1100271317309939,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}