2 ms·
Model-Based Machine Learning
- cultus 8y agoSkimming this, it mostly focuses on Bayesian probabalistic models. I think these are underutilized in machine learning tasks, especially since it can be so easy nowadays with tools like PyMC and fast variational methods. If you have a reasonable data-generating model in mind, that information should be incorporated into the prior.
- sctb 8y agoYesterday: https://news.ycombinator.com/item?id=18201986 https://news.ycombinator.com/item?id=18201986.