Summary of Double-bounded Optimal Transport For Advanced Clustering and Classification, by Liangliang Shi et al.
Double-Bounded Optimal Transport for Advanced Clustering and Classification
by Liangliang Shi, Zhaoqi Shen, Junchi Yan
First submitted to arxiv on: 21 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Optimization and Control (math.OC)
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 Optimal transport (OT) is gaining popularity in machine learning as a way to transform a source distribution into a target one at minimal cost. Traditional OT assumes predetermined source and target distributions, which isn’t practical for real-world scenarios where targets are often unknown or uncertain. This paper proposes Doubly Bounded Optimal Transport (DB-OT), an approach that restricts the target distribution within two boundaries, allowing for more flexibility in finding solutions. Three scaling-based algorithms are developed to calculate the optimal solution. DB-OT is also applied to barycenter-based clustering, helping avoid cluster concentration issues. Furthermore, the authors connect OT to classification and propose a novel inference scheme based on DB-OT, achieving improved results even with vanilla Softmax features. The paper’s contributions include developing DB-OT techniques for long-tailed classification and improving inference in the testing stage. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Optimal transport is a way to move one distribution of things into another distribution at the lowest cost. In real life, we often don’t know what the target distribution should look like, which makes it hard to use traditional methods. This paper presents a new approach called Doubly Bounded Optimal Transport that gives more freedom in finding solutions. They also show how this method can be used for grouping things together and improving classification. The authors connect optimal transport to another important concept, classification, and propose a new way of making predictions. Overall, the paper makes it easier to use optimal transport in real-world problems. |
Keywords
* Artificial intelligence * Classification * Clustering * Inference * Machine learning * Softmax
