Mathematics Data Science Seminar: Yan Wang, Prediction Done Right with Wrong Machine Learning Models
This event is in the past.
1:30 p.m. to 2:30 p.m.
Yan Wang’s data science seminar talk on 10/04
Title: Prediction Done Right with Wrong Machine Learning Models
Abstract: Conformal prediction (CP) is a technique to provide theoretically justified prediction intervals under minimal assumptions. In practice, it can be applied to almost any machine learning model: linear regression, random forests, neural networks, etc. You name it. Even if you always guess πΜ π+1 = 0 irrespective of the corresponding ππ+1 for a future test point (ππ+1, ππ+1), CP can provide a prediction interval πΆπ,πΌ(ππ+1) such that ππ+1 ∈ πΆπ,πΌ(ππ+1) with probability at least 1 − πΌ, where πΌ ∈ (0,1) is a prescribed constant and πΆπ,πΌ(ππ+1) is constructed by CP from the training set {(ππ , ππ )}π=1 π . I will introduce the basic idea of CP and some of its variants that are computationally appealing in practice. In particular, I will show that, under (really) mild conditions, by training random forests once, not only can one have a point predictor for ππ+1, but also a valid prediction interval that comes almost for free.