An RBM has visible units v and hidden units h with no within-layer connections: E(v,h) = −v·W·h − b·v − c·h. Training via Contrastive Divergence adjusts weights to make the model's distribution match the data.
—
Reconstruction Error
—
Free Energy
0
CD Steps
8
4
0.010
1.00
Top: network architecture · Bottom left: weight matrix · Bottom right: generated samples over time · Click visible units to toggle