
Episode 15: Martín Arjovsky, INRIA, on benchmarks for robustness and geometric information theory
Generally Intelligent
00:00
Nour Networks Are Compromising With Entropy
Theory parp is correct, but meaningless. Ti correct, but bakus, let's say. And the experiments, i think, just don't make much sense. The only thing that i come to complain about this paper is that it sas too many things tat it should be thee separate papers. S y basically describes how police correlations canno rise. But if you want to know which one of both of them are very well correlated with th they can e now, but the important thing in here is that a lot of times, if you just train a linear regression or a logistic classification,. a lot of time, not what i want it to always
Transcript
Play full episode