Summary of Memory-enhanced Invariant Prompt Learning For Urban Flow Prediction Under Distribution Shifts, by Haiyang Jiang and Tong Chen and Wentao Zhang and Nguyen Quoc Viet Hung and Yuan Yuan and Yong Li and Lizhen Cui
Memory-enhanced Invariant Prompt Learning for Urban Flow Prediction under Distribution Shifts
by Haiyang Jiang, Tong Chen, Wentao Zhang, Nguyen Quoc Viet Hung, Yuan Yuan, Yong Li, Lizhen Cui
First submitted to arxiv on: 7 Dec 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 A novel framework called Memory-enhanced Invariant Prompt learning (MIP) is proposed for urban flow prediction under constant distribution shifts. STGNNs have established themselves as capable predictors but tend to suffer from distribution shifts common with urban flow data. To address this, MIP uses a learnable memory bank trained to memorize causal features within the spatial-temporal graph. This allows adaptive extraction of invariant and variant prompts for each location at every time step. Instead of intervening raw data based on simulated environments, MIP directly intervenes on variant prompts across space and time. Invariant learning then minimizes prediction variance to ensure predictions are only made with invariant features. Comparative experiments on two public urban flow datasets demonstrate the robustness of MIP against OOD data. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Urban flow prediction is a way to estimate traffic flow for a given location in the future. Models like Spatial-Temporal Graph Neural Networks (STGNNs) can do this, but they don’t always work well because the urban environment changes and is hard to predict. In this paper, scientists came up with a new idea called MIP (Memory-enhanced Invariant Prompt learning). MIP helps the model remember important patterns in the data and use them to make predictions that are less affected by changes in the environment. They tested MIP on real traffic flow data and showed it worked better than other models. |
Keywords
* Artificial intelligence * Prompt
