AVSurfer123 / adversarial-cooperation

Code for the CS 285 Deep Reinforcement Learning Final Project
0 stars 0 forks source link

Implement JSMA Saliency Map Attack #2

Open shawnshact opened 3 years ago

shawnshact commented 3 years ago

Given some target action, generate a saliency map that will perturb observation input to manipulate a victim agent to take the action.

AVSurfer123 commented 3 years ago

https://discuss.pytorch.org/t/how-to-compute-jacobian-matrix-in-pytorch/14968/4

starter code: https://github.com/ast0414/adversarial-example/blob/master/craft.py

maybe faster Jacobian method: https://gist.github.com/sbarratt/37356c46ad1350d4c30aefbd488a4faa