Same Smell
One cVA puff into MaleCNS (he) and FlyWire (she) at once: he lunges, she pauses. Extra modes play Pong between the two brains and let both courtship circuits swipe dating profiles. LIF timing is stretched; bodies are puppets on real descending output.
Neurons are simulated — leaky integrate-and-fire or better — with documented, tuned parameters.
Dispute this rating →The evidence
No evidence, no level.The simulation is a leaky integrate-and-fire model with stated membrane time constant, synaptic decay, threshold, reset and refractory period.
https://github.com/anzal1/samesmell/blob/f5891d9206802e172c619e2bcede9d25e2cb79f1/src/sim.js#L12-L18The per-step membrane update integrates leak, synaptic current and an intrinsic adaptation term; no synaptic weight appears on the left-hand side of any assignment in the loop.
https://github.com/anzal1/samesmell/blob/f5891d9206802e172c619e2bcede9d25e2cb79f1/src/sim.js#L347-L353Synaptic weights are set once when the graph is built, from fixed spine/feedforward/lateral budget constants, rather than from the connectome's synapse counts.
https://github.com/anzal1/samesmell/blob/f5891d9206802e172c619e2bcede9d25e2cb79f1/src/sim.js#L147-L165The app loads the real traced connectome subsets first and only falls back to a procedural graph if they are unavailable.
https://github.com/anzal1/samesmell/blob/f5891d9206802e172c619e2bcede9d25e2cb79f1/src/main.js#L130-L141
Real MaleCNS v1.0 and FlyWire subsets (615 neurons / 5,837 synapse-counted edges on the male side, fetched from neuPrint) are driven through a leaky integrate-and-fire simulation with declared membrane and synaptic time constants, threshold, reset and refractory period, so the flies' reactions are produced by neuron dynamics rather than by keyframes. Not L2: every synaptic weight is assigned once at graph-build time from fixed budget constants and never touched again inside the integration loop, and the only state that adapts during a run is a per-neuron spike-frequency adaptation current, which is intrinsic excitability, not a synaptic change. Worth noting for a reader that the connectome supplies the topology but not the weights here - the README lists synapse-count-weighted connections and habituation as future work, not as shipped behaviour.
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.