Self-Evolving Neural Network
This is AiA's brain, running live in your browser. It learns from every sample, remembers its experience, adapts its learning rate, and when it gets stuck it evolves its own architecture - growing new neurons. You are watching a network improve itself in real time. (Python twin: core/neural/ in the AiA repo.)
Accuracy
0%
Loss
—
Generation
0
Samples
0
Neurons
0
Evolutions
0
Prediction map · click the map to teach it
The brain never pauses and never gets force-evolved - it is autonomous. You can challenge it (outlier), energize it (rebirth), preserve it (export/import genome) or benchmark it (read-only). Nothing on this page can destroy learned knowledge. Blue cells = predicted 1 · purple = predicted 0 · colored dots are training samples (blue = 1, red = 0). Click a cell to drop your own teaching sample.
Loss curve
One bar = one neuron · numbers below = neurons per layer · the architecture changes only when the network evolves.