The paper introduces a general framework for approximating two-layer neural networks using sparse Mixtures of Experts (MoEs) and product-key memories (PKMs). The proposed methods improve both MoEs and PKMs, showing that they are competitive with dense models while being more resource efficient.

https://arxiv.org/abs//2310.10837

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers