Archive MaleCNS research sources, provenance, bibliography and immutable data inventory

2026-09-29 03:38:32 +08:00
commit 3981628ff8
682 changed files with 161475 additions and 0 deletions
@@ -0,0 +1,41 @@
+phase: 4U
+experiment_root: /data/digital-drosophila/experiments/phase4u-odor-generalization
+model_root: /data/digital-drosophila/models/male-cns-lif/odor-timing-generalization-v1
+runtime_root: /data/digital-drosophila/models/male-cns-lif/deterministic-runtime-v2
+phase4p_root: /data/digital-drosophila/experiments/phase4p-deterministic-runtime
+prior_root: /data/digital-drosophila/experiments/phase4t-timing-stability
+prior_model: /data/digital-drosophila/models/male-cns-lif/timing-stability-v1
+duration_ms: 120
+target_compartment: "y5B'2a"
+matched_DAN: PAM15
+candidate_US_ms: [10, 15, 20, 25, 30, 40, 60, 90]
+windows_ms: [5, 10, 20, 50, 100, 250]
+calibration_seeds: [118101,118102,118103,118104,118105,118106,118107,118108,118109,118110,118111,118112,118113,118114,118115,118116,118117,118118,118119,118120]
+validation_seeds: [119101,119102,119103,119104,119105,119106,119107,119108,119109,119110,119111,119112,119113,119114,119115,119116,119117,119118,119119,119120]
+odor_seed_offset: 2000000
+seed_namespace_offset: 50000000
+panel:
+  calibration: all six frozen Phase4L calibration odors
+  novel_count: 6
+  exclusion: all old twelve odor IDs and names
+  quality_filter: same minimum five strong mapped glomeruli and mapped_response>0 as frozen encoder; mapped_signal_fraction at least 0.9
+  selection: stable sort eligible new odors by mapped_response_sum then odor_id; six equal rank bins; choose lower median rank of each bin
+  no_selection_on_runtime_activity: true
+timing_gate:
+  ratio: 3.0
+  qualifying_fraction_each_odor: 0.8
+  mean_ratio_each_odor: 3.0
+  selection: maximize minimum across six calibration odors of q20 per-seed ratio; ties shortest window then earliest US
+  random: exact uniform expectation over 1200 timestep gates on same KC events
+  unchanged_from_4T: true
+  on_no_qualifying_calibration: stop; do not access new odor dynamic trials or choose a least-bad parameter
+live_US_checks:
+  seeds_per_validation_odor: 1
+  random_seed_offset: 29000000
+scope:
+  weight_updates: false
+  formal_AB_used: false
+  per_odor_US_or_event_triggered_US: false
+  change_encoder_rates_dynamics_or_signs: false
+  new_candidate_requires_all_validation_odors_pass: true
+  gate_scope: engineering fixed-US timing, not biological memory