This repo contains the source code of Testability-Aware Low Power Controller Design with Evolutionary Learning, ITC 2021.
The entry to the core algorithm is ga.py. Other files are not related to the paper and can be ignored.
Created on 6/11/2020. The repo has been updated again on 4/09/2020. Try BNN for lossless testing compression. This repo only contains only a rough implementation of a binarized auto-encoder for compressing the test cubes.
The codes are referred from jiecaoyu/XNOR-Net-PyTorch
- From high-level pespective, they are the same, as BNN can be seem as a stacted XOR Net structure where its parameters should be learned from data.
- 1-layer decoder of BNN is exactly a XOR network.
- Using GA to search an optimal XOR matrix for EDT, which are more effective than random XOR matrix.
How to initialize the XORNet is important. Usually, we need the matrix to be orthogonal. And we might refer to this Xavier Initialization paper.