Help the Fly Escape
Arrange household objects and observe flies navigating a 3D browser game. Rust/WASM simulation using a selected MaleCNS circuit.
Neurons are simulated — leaky integrate-and-fire or better — with documented, tuned parameters.
Dispute this rating →The evidence
No evidence, no level.LIF parameters are declared with defaults of tau 20, threshold 1.0, reset 0.0 and a 2-step refractory period.
https://github.com/dzhng/fly-escape/blob/06540138d55127f87e69f8cca9982c4f1328be90/crates/sim/src/lif.rs#L8-L26Each step integrates the membrane voltage from leak, synaptic input, external drive and noise, then spikes, resets and enters refractory.
https://github.com/dzhng/fly-escape/blob/06540138d55127f87e69f8cca9982c4f1328be90/crates/sim/src/lif.rs#L291-L312Synaptic weights live on the loaded graph and are only read when accumulating input - nothing in the simulation writes them.
https://github.com/dzhng/fly-escape/blob/06540138d55127f87e69f8cca9982c4f1328be90/crates/sim/src/lif.rs#L276-L282
The Rust simulation crate integrates a leaky integrate-and-fire population with stated parameters (tau 20, threshold 1.0, reset 0.0, refractory 2 steps) plus baseline drive and seeded noise, voltages resetting and entering refractory on spike, with per-fly independent state. Documented neuron dynamics on a real MaleCNS subgraph, so L1. Not L2: the graph's weight vector is built once when the prepared subgraph is loaded and is only ever indexed to accumulate synaptic input; there is no plasticity, dopamine, STDP, eligibility trace or weight write anywhere in crates/sim/src. The README says as much in its own words - the connections come from real data, while the neuron equations, sensory inputs and movement rules are models.
Record
Nearby
All projects →MaleCNS simulation connected to ViZDoom through modeled visual inputs and a fixed button readout. Includes plasticity experiments and negative validation results; learned survival has not been demonstrated.
Research framework comparing a MaleCNS-constrained sparse recurrent controller with rewired graphs, conventional neural networks, and other controls on simple VizDoom tasks.
Ratings on thelearningfly.com are proposals, not verdicts. Every one of them is arguable in public.