Beyond Empirical Risk Minimization: the lessons of deep learning

[@JasminUwU]

Very high quality talk, thank you very much!

[@egorpanfilov]

Great talk! Thank you Mikhail for wonderful explanations!

[@selfhelp119]

Good presentation! thank you for sharing.

[@straumits]

Fascinating!

[@jnaniify]

thx - great talk!
but... random is ~90% test error (~depending on the distribution of the random guess) and not anything above 90 - being reliably 100% wrong is not random, but rather accurate misclassification, right?