In the previous posts, I explained the covariate shift but missed the most important part… How can we detect it? This simple diagram is breaking down various covariate shift detection methods: 1) Multivariate - detects a shift in a joint(multiple features) distribution 2) Univariate - detects changes in a single feature distribution, using either: ↳ Statistical tests ↳ Distance methods Each of these methods has trade-offs based on the type of shift, but I will talk more about it in the next post👀
