The blog has been eerily silent for most of 2018, here is why:
- The main culprit is definitely the COLT 2018 chairing. This year we received a surprise 50% increase in number of submissions. This is great news for the community, and it led to a truly fantastic program. We are looking forward to see many of you in Stockholm! Note that the early registration deadline is tomorrow (May 30th).
- The first volume of our new Mathematical Statistics and Learning journal just got published. If you care about being evaluated by top experts, and/or get the visibility of being in a highly selective mathematical journal, you should consider submitting your highest quality work there!
- The first set of video lectures on the youtube channel is now complete. I realize that the quality is not perfect and that it is hard to read the blackboard. I will try to improve this for the future videos (which should be posted some time during the summer).
- Finally I am quite excited to have a first preprint (joint work with Eric Price and Ilya Razenshteyn) at least loosely related to deep learning. It is titled “Adversarial examples from computational constraints”, which sums up the paper pretty well. In a nutshell, we prove that avoiding adversarial examples is computationally hard, even in situations where there exists a very robust classifier. We are looking forward to keep the deep learning community’s feedback on this work!