Loading Now

Summary of Representation Collapsing Problems in Vector Quantization, by Wenhao Zhao et al.


Representation Collapsing Problems in Vector Quantization

by Wenhao Zhao, Qiran Zou, Rushi Shah, Dianbo Liu

First submitted to arxiv on: 25 Nov 2024

Categories

  • Main: Machine Learning (cs.LG)
  • Secondary: Artificial Intelligence (cs.AI)

     Abstract of paper      PDF of paper


GrooveSquid.com Paper Summaries

GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!

Summary difficulty Written by Summary
High Paper authors High Difficulty Summary
Read the original abstract here
Medium GrooveSquid.com (original content) Medium Difficulty Summary
The paper investigates the phenomenon of “representation collapse” in vector quantization, a technique used to discretize continuous representations in machine learning models. Specifically, it focuses on how codebook tokens or latent embeddings lose their discriminative power and converge to a limited subset of values, compromising the model’s ability to capture diverse data patterns. The study uses both synthetic and real datasets to identify the severity of different types of collapses and their triggering conditions.
Low GrooveSquid.com (original content) Low Difficulty Summary
Vector quantization is used in large language models, diffusion models, and other generative models to tokenize data representations. However, it remains unclear how vector quantization behaves in these models. This paper explores this issue by looking at “representation collapse” – when codebook tokens or latent embeddings lose their ability to capture different patterns in the data.

Keywords

» Artificial intelligence  » Machine learning  » Quantization