This is a toy implementation of a boltzman restricted machine.
You can also add more "images" in images.py
to train the network.
See my blogpost for more details.
Run ./main.py for a while so network learns from images.py
.
Then you can run read.py
to evaluate an image against network, or
run dream.py
to see how the network fluctuates around it's "fantasies".
You can also do cat img-2.img | ./read.py
to check how network
reflects on the image supplied.