Recall Result
Run recall to see the original, corrupted, and reconstructed patterns.
Stored Patterns
show ↓hide ↑Loading stored patterns...
How It Works
01
Hebbian autoassociative memory stores each bipolar pattern as a stable attractor of a recurrent neural network.
02
The weight matrix is built from outer products of the stored patterns and its diagonal is cleared to avoid self-reinforcement.
03
During recall, the network iteratively updates each cell until the state converges. Noise flips cells, masking removes information, and the network fills in the missing structure from learned correlations.