How jxsh341/project-icarus works
fly-cognition is a connectome-derived neural research model for studying innate behavior, associative learning, retention, and forgetting. Its principal workflow is to obtain and transform connectome records, construct signed connectivity, simulate neural activity, and train or modify synaptic weights; experiments then evaluate behavioral and substrate-level outcomes. The repository also documents a closed-loop embodiment scaffold and distinct forgetting-certification and synaptic-erasure studies. Some experimental subsystems are represented by their exact scripts or directories because their implementations were not sampled; links without direct evidence are omitted.
Main entry points: training.py Training driver · sensorimotor_loop.py Sensorimotor loop · loader.py Snapshot loader · lif.py LIF simulation
Components 20
Connectome data
- Snapshot loader
fly/data/loader.py - Data schemas
fly/data/schemas.py - Provenance
fly/data/provenance.py - Record transform
fly/data/transform.py - Neuron records
fly/neurons/models.py - Synapse records
fly/synapses/models.py
Neural substrate
- Signed connectivity
fly/network/graph.py - LIF simulation
fly/dynamics/lif.py - Substrate API
fly/api.py
Learning and embodiment
- Plasticity rules
fly/plasticity.py - Training driver
fly/training.py - Body and sensors
fly/embodiment/body_graph.py - Sensorimotor loop
fly/embodiment/sensorimotor_loop.py
Research experiments
- Acquisition studies
scripts/phase2_wp3_pilot.py - Eligibility and sign
scripts/phase4_eligibility_sign_rule.py - Extinction analysis
scripts/phase5_expA_analysis.py - Forgetting oracle
scripts/phase6_l6a_oracle.py - Targeted erasure
scripts/phase6_wp_l6b_rankforget.py - Closed-loop study
scripts/phase7e1_harness.py
Other components
- neuPrint
Connections 15
- neuPrint → Snapshot loader: supplies snapshots
README.md - Snapshot loader → Neuron records: loads records
fly/data/loader.py - Snapshot loader → Synapse records: loads records
fly/data/loader.py - Neuron records → Signed connectivity: provides neurons
fly/network/graph.py - Synapse records → Signed connectivity: provides synapses
fly/network/graph.py - Substrate API → LIF simulation: simulates trials
fly/api.py - Training driver → LIF simulation: runs trials
fly/training.py - Training driver → Plasticity rules: applies updates
fly/training.py - Sensorimotor loop → Body and sensors: reads sensors
fly/embodiment/sensorimotor_loop.py - Sensorimotor loop → LIF simulation: simulates step
fly/embodiment/sensorimotor_loop.py - Acquisition studies → Training driver: runs learning
scripts/phase2_wp3_pilot.py - Eligibility and sign → Training driver: configures unlearning
- Extinction analysis → Forgetting oracle: feeds evaluation
scripts/phase6_l6a_oracle.py - Forgetting oracle → Targeted erasure: compares erasure
- Sensorimotor loop → Closed-loop study: supports closed-loop study