Summary of Propensity Score Alignment Of Unpaired Multimodal Data, by Johnny Xi et al.
Propensity Score Alignment of Unpaired Multimodal Data
by Johnny Xi, Jana Osea, Zuheng Xu, Jason Hartford
First submitted to arxiv on: 2 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Methodology (stat.ME); 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 This paper addresses the challenge of aligning unpaired samples across disparate modalities in multimodal representation learning, particularly in fields like biology where measurement devices often destroy samples. By drawing an analogy between potential outcomes in causal inference and potential views in multimodal observations, the authors use Rubin’s framework to estimate a common space for matching samples. They assume experimentally perturbed samples are collected, estimating a propensity score from each modality that encapsulates shared information between a latent state and treatment. This allows defining a distance between samples, which is used to leverage two alignment techniques: shared nearest neighbours (SNN) and optimal transport (OT) matching. Experimental results show OT matching achieves significant improvements over state-of-the-art approaches in both synthetic and real-world datasets from the NeurIPS Multimodal Single-Cell Integration Challenge. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps computers learn about different kinds of data, like images and genes, even when they’re not paired together. This is a big problem because sometimes we can’t get paired samples in fields like biology. The authors come up with a new way to match unpaired samples by using a framework that’s usually used for understanding cause-and-effect relationships. They show how their approach works better than others in both fake and real-world datasets. |
Keywords
* Artificial intelligence * Alignment * Inference * Representation learning
