Summary of Scaling Laws For Data Filtering — Data Curation Cannot Be Compute Agnostic, by Sachin Goyal et al.
Scaling Laws for Data Filtering – Data Curation cannot be Compute Agnostic
by Sachin Goyal, Pratyush Maini, Zachary C. Lipton, Aditi Raghunathan, J. Zico Kolter
First submitted to arxiv on: 10 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 This paper presents a novel approach to data curation for vision-language models (VLMs). Traditional methods involve filtering out low-quality data, but this often results in suboptimal performance. The authors introduce neural scaling laws that account for the varying utility of web data and its diminishing returns with repetition. These laws enable the estimation of model performance on combined datasets without joint training. The key finding is that data curation must consider the total compute budget for a model, allowing for the creation of a pareto-frontier for optimal data selection. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you’re trying to build a really good computer program that can understand and generate images. You need lots of pictures and text to train it, but not all of them are useful. This paper shows how to choose the best “good” pictures and texts to use, so your program can learn quickly and well. The authors also explain why choosing just the “best” data isn’t enough – you need to consider how much computer power you have to process that data too. |
Keywords
» Artificial intelligence » Scaling laws
