Summary of Energy-based Model For Accurate Shapley Value Estimation in Interpretable Deep Learning Predictive Modeling, by Cheng Lu et al.
Energy-based Model for Accurate Shapley Value Estimation in Interpretable Deep Learning Predictive Modeling
by Cheng Lu, Jiusun Zeng, Yu Xia, Jinhui Cai, Shihua Luo
First submitted to arxiv on: 1 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 Shapley value-based explainable artificial intelligence (XAI) has been a popular tool for interpreting deep learning models. However, estimating Shapley values accurately and efficiently is challenging due to the exponential growth in computation load with input features. While existing accelerated methods compromise on estimation accuracy for efficiency, we present EmSHAP, an energy-based model that estimates the expectation of Shapley contribution functions under arbitrary subsets of features given the rest. This involves energy networks for approximating unnormalized conditional densities and GRU (Gated Recurrent Unit) networks for approximating partition functions, which eliminate input ordering impacts. To evaluate performance, we analyze error bounds of EmSHAP alongside KernelSHAP and VAEAC using Theorems 1-3, showing that EmSHAP has tighter error bounds than the state-of-the-art methods. Case studies on two applications demonstrate enhanced estimation accuracy for EmSHAP. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary EmSHAP is a new way to make artificial intelligence more understandable. Right now, it’s hard to figure out why AI models are making certain predictions because there are too many inputs and calculations to keep track of. Existing solutions sacrifice some accuracy in order to work faster, but EmSHAP tries to do both well. It uses special networks called energy networks and GRU (Gated Recurrent Unit) networks that help it understand what’s important and what’s not. This makes it better at explaining its predictions than other methods. We tested EmSHAP on two real-world problems and found that it does a great job of understanding how the AI model is working. |
Keywords
* Artificial intelligence * Deep learning * Energy based model
