1 paper · 1 filter
Yilun Kuang, Yash Dagade, Deep Chakraborty +4
Self-supervised learning aims to learn maximally informative representations, but explicit information maximization is hindered by the curse of dimensionality. Existing methods lik…