Summary of Mbexplainer: Multilevel Bandit-based Explanations For Downstream Models with Augmented Graph Embeddings, by Ashkan Golgoon et al.
MBExplainer: Multilevel bandit-based explanations for downstream models with augmented graph embeddings
by Ashkan Golgoon, Ryan Franks, Khashayar Filom, Arjun Ravi Kannan
First submitted to arxiv on: 1 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); Numerical Analysis (math.NA); Machine Learning (stat.ML)
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 proposed approach, MBExplainer, is a model-agnostic explanation method for ensemble models that combine graph neural network (GNN) embeddings with tabular features. The goal is to explain the output of these models by identifying the most important subgraphs, nodal features, and augmented downstream features contributing to an instance prediction. MBExplainer uses a game-theoretic formulation to assign Shapley values to each component and iteratively prunes local search spaces using Monte Carlo Tree Search. The approach also incorporates contextual bandits for efficient budget allocation. Numerical examples on public graph datasets demonstrate the effectiveness of MBExplainer for node and graph classification tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary MBExplainer is a tool that helps us understand how predictions are made by combining information from graphs with other data. It works with models that use learned patterns in these graphs, as well as other facts. The goal is to provide simple explanations for what makes a prediction happen. MBExplainer does this by identifying the most important parts of the graph and the features used to make the prediction. It uses special algorithms to find the right answers quickly. This approach was tested on real-world data sets and showed it can work well for predicting things about nodes in graphs. |
Keywords
» Artificial intelligence » Classification » Gnn » Graph neural network
