Evaluating models without test data (Practical AI #194)
Changelog Master Feed
00:00
Deep Learning - What Is It About Alpha?
The way neural networks learn is they learn the multi fractal nature of the data. That's why they work so well on things like text and images, but not tabular data sets. So what we're doing is there are correlations in the data data is correlated. And frequently you're trying to learn very subtle correlations that couldn't be found using some simple clustering algorithm or an SVM. The regular is or failed test can detect difference between when there's a problem with the optimizer and when there's actual natural structure in the data.
Transcript
Play full episode
