Self-training with unlabeled data to improve generalization

Mar 10, 2022

A very interesting theoretical paper that shows why self-training with unlabeled data improves generalization performance.

Another good thing about this paper is that they used a 2-layer network (just one hidden layer). This work shows that you can do very interesting and profound research using limited resources. ?

HT: Taesun Lyu on TensorFlow KR.

https://arxiv.org/pdf/2201.08514.pdf

← Back to all articles         Quick Navigation:    Next:[ j ] – Prev:[ k ] – List:[ l ]