This repository serves as supplementary for the paper:
A mean field theory of quantized deep networks: A Mean Field Theory of Quantized Deep Networks: The Quantization-Depth Trade-Off.
And includes several tools to assist research of deep-neural-networks using statistical physics
- Python 3.0
- Pytorch with Cuda (Tested with 1.0.1.post2)
- Jupyter (Lab)
- Standard python packages, including: numpy, matplotlib, scipy
No installation required.
The python files in "./src" includes several tools usage throughout the research.
The jupyter-examples include examples, which utilizes several of those tools, and allow reproduction of the main figures in the paper.