Summary of Stochastic Amortization: a Unified Approach to Accelerate Feature and Data Attribution, by Ian Covert et al.
Stochastic Amortization: A Unified Approach to Accelerate Feature and Data Attribution
by Ian Covert, Chanwoo Kim, Su-In Lee, James Zou, Tatsunori Hashimoto
First submitted to arxiv on: 29 Jan 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 tackles the challenge of efficient explainable machine learning (XAI) for large datasets. Many XAI tasks, such as data valuation and feature attribution, rely on expensive computations for each data point. To address this issue, researchers have explored amortizing the process by training a network to directly predict the desired output. However, training such models with exact labels is often infeasible. The proposed solution involves training these amortized models with noisy labels, which surprisingly proves effective. Through theoretical analysis and experiments with various models and datasets, the authors demonstrate that this approach can tolerate high noise levels and achieve significant speedups for feature attribution and data valuation methods, often resulting in an order of magnitude acceleration over existing approaches. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make machine learning more understandable. When we try to understand how machines are making decisions, it takes a lot of computer power. Scientists found a way to use less computer power by using noisy labels instead of exact labels. This works really well and can speed up many tasks by up to 10 times. The researchers did some math problems and tested their idea with different models and data sets. |
