3 ms·
Ask HN: What are some good but forgotten ideas in ML?
- probinso 6y agoManifold learning is not very common anymore
- probinso 6y agoI would also like to see a lot more probabilistic graphical models/ Bayesian networks.
- nikonyrh 6y agoIs feature engineering still a thing? For example transforming circular images (like coins) to log-polar coordinates and maybe even taking the absolute value of Fourier transform to get rotation-invariant features.