17
Reconstructions
by model trained
on 2ís
Data
Reconstructions
by model trained
on 3ís
100 hidden units
(features)
j
j
256 visible
units (pixels)
i
i
reconstruction
data