Plenty of Nothing

I occasionally archive my grandfather’s unpublished poems here (in traditional chinese). Most of them were burned during the cultural revolution. 寄蔣介石 To Chiang Kai-Shek 革命追隨共甘苦。一旦生離做叛臣。 不及翠村酸棗樹。...

Two true things in the world: learning comes from compression, and models should be sparse. In this post, we will talk about the latter, focusing on the training techniques for sparse models. 0....

This is an additional note to my previous note. Also see the new paper list on IDS, and the note on Andreas Krause’s IMSI talk. 1. High-level Idea Decoupling Exploitation and Exploration \begin...