SOCRATES is a unified platform for neural network analysis developed by Sun Jun’s team at SMU.
Unlike most existing neural network analysis approaches which are scattered (i.e., each approach tackles some restricted classes of neural networks against certain particular properties), incomparable (i.e., each approach has its own assumptions and input format) and thus hard to apply, reuse or extend, SOCRATES aims at providing a unified platform for neural network testing, verification and repair. Specifically, it supports a standardized format for a variety of neural network models, an asseration language for property specification as well as many engines for testing, verifying, and repairing neural network models.
SOCRATES is still in active development. Any suggestions and collaborations are welcomed.
The benchmark of SOCRATES is available at benchmark. More information are available at SOCRATES website.
virtualenv -p python3 socrates_venv
source socrates_venv/bin/activate
pip install numpy scipy matplotlib torch autograd antlr4-python3-runtime==4.8 sklearn pyswarms gensim python-Levenshtein
-
SOCRATES: Towards a Unified Platform for Neural Network Verification
Long H. Pham, Jiaying Li, Jun Sun
- You may use GitHub system to raise new issues or suggestions.
- For collaborations, you may reach us via email addresses.