This page contains code useful to reproduce some of the experiments reported in:
M. Ranzato, V. Mnih, G. Hinton,
"Generating More
Realistic Images Using Gated MRF's". Advances in Neural Information
Processing Systems 23 (NIPS 2010), Vancouver, 2010.
mPoT on image patches
Download
here
the CUDAMAT version
which requires Volodymyr Mnih's
CUDAMAT
and a GPU board to run.
Download
here
the GNUMPY version
which is a bit slower but easier to read.
This runs both on GPU and CPU.
Tiled mPoT on high-resolution images
This is the code to train TConv mPoT on high-resolution images (see fig.3-A of the paper).