MushroomBody_Pong v1

Synaptic weights for the Pong game of BeatTheFly -- A Smart Fruit Fly is playing Pong against you: a spiking network wired as the real Drosophila mushroom body connectome that moves a Pong paddle in real time.

The anatomical connectome gives you wiring, not synaptic strengths. Ours are trained.

Synaptic weights trained with PHCSSM parallel-scan mode, deployment in sequential RSNN mode (PHCSSM).

made by Po-Han Chiang @ NYCU

What the mushroom body does in the fly

The mushroom body is the fly's centre for associative learning and memory. Its gate is dopaminergic — 340 dopaminergic neurons over 3,160 real synapses — and what they write into the fast weight fades within about a frame and a half.

Architecture

• Wiring: MaleCNS v1.0 mushroom body -- 4,510 neurons (KC 4,064; MBON 97; DAN 340; APL 2; other 7) and 1,027,152 neuron-to-neuron connections. The connectivity mask is fixed to the connectome; 732,618 connections carry a nonzero weight and 0 weights lie off the connectome.
• Dale's law: one sign per presynaptic neuron from predicted neurotransmitters (excitatory 4,114, inhibitory 52, modulatory 344); 0 weights violate it.
• Input: each frame provides 6 numbers seen from the fly's side of the court: the ball's position and velocity and the positions of both paddles. A linear encoder with LayerNorm drives only the input population: 4,064 Kenyon cells.
• Neurons: leaky integrate-and-fire with per-neuron leak, threshold and reset; synaptic delay of one step.
• Readout: linear map from the membrane voltage of the output population only, 97 MBON, to 3 paddle commands (stay, up, down).
• Neuromodulatory gate: 340 of this region's own known-neuromodulator cells (DAN (dopaminergic)) gate the current into the output population, masked to the 3,160 real gate-cell -> output synapses in the connectome.
• Fast weight: a value written onto the 61,210 real input -> output synapses, gated by the same cells, read back into the output neurons' voltages. It is read with the current frame's input spikes and written with the same frame's, and it decays: one trainable factor per output neuron, a decay time constant of about 1.41 frames (23.5 ms at 60 fps).
• Deployment: sequential RSNN mode, one timestep per frame at 60 frames per second, with the neuron state carried across the whole game.

Data sources

Data source: actions of a scripted Pong player.

Evaluation

20 games to 11 points against each scripted player, the fly playing its top command every frame:

opponent	games won	point share	fly's return rate
a noisy scripted player	10 / 20	0.506	87.0%
a weak scripted player	20 / 20	0.827	87.4%
a perfect scripted player	0 / 20	0.000	89.5%

On 131,072 held-out frames the fly's choice matches the scripted player's action on 97.0% in sequential RSNN mode, and the parallel-scan and sequential modes choose the same action on 99.69% of frames.

Files

• manifest.json -- every tensor (file, dtype, shape, bytes), the model scalars and a connectome audit.
• info.json -- neuron metadata used by the page (cell classes, hemispheres, soma coordinates).
• selfcheck_<precision>.json -- reference observations and logits that the page replays when it loads.
• fp16/, fp32/ -- raw little-endian arrays.

Two precisions are listed in the manifest: fp16w32 (default, 5.4 MB: float16 for the readout matrix dec_w, float32 for the recurrent weights and all other tensors) and fp16 (4.0 MB, recurrent weights in float16 as well).

The recurrent weight matrix W[dst, src] is stored in CSC order by source neuron (W_colptr, W_rowidx, W_vals): each step multiplies W by a sparse binary spike vector, so the engine visits only the columns of the neurons that spiked. in_idx lists the input population and out_idx the output population. enc_obs_T [6, H] is the observation encoder and dec_w [3, 97] is read against the output population's voltage.

name	file	dtype	shape
enc_obs_T	fp32/enc_obs_T.bin	float32	6x4510
enc_obs_b	fp32/enc_obs_b.bin	float32	4510
ln_obs_w	fp32/ln_obs_w.bin	float32	4510
ln_obs_b	fp32/ln_obs_b.bin	float32	4510
teach_T	fp32/teach_T.bin	float32	6x340
teach_b	fp32/teach_b.bin	float32	340
W_gate	fp32/W_gate.bin	float32	97x340
W_val	fp32/W_val.bin	float32	97x340
gate_idx	fp32/gate_idx.bin	int32	340
fw_in	fp32/fw_in.bin	int32	61210
fw_out	fp32/fw_out.bin	int32	61210
gamma	fp32/gamma.bin	float32	97
dec_w	fp16/dec_w.bin	float16	3x97
dec_b	fp32/dec_b.bin	float32	3
alpha_exc	fp32/alpha_exc.bin	float32	4510
alpha_inh	fp32/alpha_inh.bin	float32	4510
v_th	fp32/v_th.bin	float32	4510
reset_weight	fp32/reset_weight.bin	float32	4510
in_idx	fp32/in_idx.bin	int32	4064
out_idx	fp32/out_idx.bin	int32	97
W_colptr	fp32/W_colptr.bin	uint32	4511
W_rowidx	fp32/W_rowidx.bin	uint16	732618
W_vals	fp32/W_vals.bin	float32	732618

Numerical check: 0 of 41,708,480 spike bits differ from the reference on the same weights (2,048 held-out frames and a 7,200-frame closed-loop game).

Limitations

The fly learned by copying a scripted player: it gets no reward and does not plan ahead. It has been tested only in this simulator and only against scripted players, and it loses every game to a perfect one.

License and attribution

Weights: CC-BY-NC-4.0. They are derived from the MaleCNS v1.0 connectome (Janelia FlyEM and collaborators, https://male-cns.janelia.org/, CC-BY-4.0) and trained with PHCSSM (https://arxiv.org/abs/2604.01295); please credit both.

Citation

PHCSSM: https://arxiv.org/abs/2604.01295