1
Your ratings — visible layer
v in [0, 1]
0
Rated
0
To predict
—
Avg rating
Rate the movies you've seen. Any unrated ones get predicted by the model.
Weight matrix W
Each value shows how strongly a movie connects to a hidden taste factor. Larger = stronger link.
Forward
pass
pass
Hidden
layer
layer
Backward
pass
pass
Weight
update
update
Results
Rate some movies above, then hit Run the RBM to see how it works step by step.
3
Backward pass — reconstructing v using h
4
Weight update — closing the gap between reality and reconstruction
5
Predictions — what you'd probably rate these
So what just happened?