Summary of Multi-modal Causal Structure Learning and Root Cause Analysis, by Lecheng Zheng et al.
Multi-modal Causal Structure Learning and Root Cause Analysis
by Lecheng Zheng, Zhengzhang Chen, Jingrui He, Haifeng Chen
First submitted to arxiv on: 4 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Methodology (stat.ME)
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 In this paper, researchers propose Mulan, a new method for performing root cause analysis (RCA) in complex systems. The approach uses multi-modal data from different sources to construct a unified causal graph, which can more accurately identify the root causes of system failures. The method involves learning log representations using a language model and then extracting modality-invariant and modality-specific features through contrastive learning. A key performance indicator-aware attention mechanism is also introduced to assess the reliability of each modality and learn a final causal graph. Finally, random walk with restart is used to simulate system fault propagation and identify potential root causes. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Root cause analysis (RCA) is important for fixing problems in big systems quickly and minimizing losses. Usually, RCA methods use data from one source only, which can be limited. Researchers are proposing a new method called Mulan that uses data from multiple sources to find the root causes of problems. They’re using special language models to learn patterns in log data, and then using those patterns to figure out how different parts of the system work together. This helps them create a map of the system that shows where problems might be coming from. They also have a way to check which sources of information are most reliable. Finally, they’re testing this method on real-world data sets and showing it works well. |
Keywords
* Artificial intelligence * Attention * Language model * Multi modal
