Geoffrey Negiar

Related to curriculum training: make examples easier at first by minimizing over input space. Try to do so especially for hard examples.

[It has been shown](https://arxiv.org/abs/2003.01690) that various improvements on textbook optimization methods allow for notable gains in attack performance. This should be implemented either in the `Attacker` class or directly in...

Parallelization of models / algorithms across several GPUs.

Similar to the Carlini & Wagner attack, we'd like the alpha to be defined datapoint wise in the penalty constraints. Ideally, given a datapoint we should be able to perform...

Cf https://github.com/openopt/copt/blob/master/examples/pytorch/adversarial_example.py

We should be able to compute operators such as prox and LMO for only a subset of the batch. For example, when performing back-tracking line-search, we only want to update...