CATBRAINThe graph
65 regional nodes. Connections derived from tract-tracing studies. The downloaded graph contains node IDs, not individual neuron recordings.
02 / RESEARCHPUBLIC CONNECTIVITY DATA · EXPERIMENTAL MODEL
A feline network.
A computational
starting point.
CATBRAIN uses a publicly available graph of connections between cat cortical regions as the fixed structure of an experimental computational model.
65CORTICAL REGIONS
1,139DIRECTED CONNECTIONS
08RESERVOIR UPDATES
01TRAINED READOUT
From connectivity
to computation.
Explore the model ↗The reservoir
Five normalized market features drive fixed input weights. Synthetic activity propagates over the published regional connectivity for eight updates.
hᵢ′ = tanh(Wᵢx + 0.65 · mean(hⱼ))
The readout
A logistic output layer is fitted to historical examples. Its target is whether the next available record closes lower.
s = σ(b + wᵀh)
BUNDLED EXPERIMENT / V0.1Evidence.
Evidence.
With its limits.
The prototype's evaluation is deliberately visible. A small historical experiment is a starting point for research.
Run the computation yourself ↗Training examples227
Chronological holdout98
Excluded: insufficient history675
Holdout Brier score · lower is better0.1092
Constant baseline Brier score0.1119
One-day sample; selective price coverage. The score is uncalibrated, the target horizon varies, and this experiment does not establish trading performance.
What the model
actually means.
Is this a real cat brain? −
It uses a published graph of regional connectivity. The computational states are artificial. It is not a biological brain simulation, an emulation of individual neurons, or a reconstruction of feline cognition.
Which part learns? +
The logistic readout weights and bias are trained. The connectivity, reservoir rule and input weights are fixed. The website lab runs this bundled model; it does not retrain it or fetch live market prices.
What do the animations represent? +
The lab's graph uses the actual 65 nodes and 1,139 links. Its layout is abstract, not anatomical. Node colors show computed model activations. The cat artwork and other decorative particles express the project's identity.