fly-mario
Full MaleCNS v1.0 playing NES Super Mario Bros with CUDA LIF dynamics, dopaminergic learning, and a live 3D web dashboard.
Neurons are simulated — leaky integrate-and-fire or better — with documented, tuned parameters.
Dispute this rating →The evidence
No evidence, no level.166,700 neurons and 25.5M connections from MaleCNS v1.0 are simulated with leaky integrate-and-fire dynamics.
https://github.com/ksanjeev284/fly-mario/blob/e5aed4d43aef9e1fc9e58728feada1c3ef473cbd/README.mdThe author states the biological wiring remains strictly frozen and only a sparse linear readout layer learns; the 'dopaminergic' mistake buffer refits those readout weights, not connectome synapses.
https://github.com/ksanjeev284/fly-mario/blob/e5aed4d43aef9e1fc9e58728feada1c3ef473cbd/README.md
A good example of why the L1/L2 line is where it is. The project describes 'dopaminergic mistake learning', but the dopaminergic weighting is applied to the trained linear readout, not to any synapse in the fly. Its own README says the biological wiring remains strictly frozen. L1 stands on the documented LIF dynamics over the full MaleCNS graph.
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.