Summary of Benchmarking Federated Learning For Semantic Datasets: Federated Scene Graph Generation, by Seungbum Ha et al.
Benchmarking Federated Learning for Semantic Datasets: Federated Scene Graph Generation
by SeungBum Ha, Taehwan Lee, Jiyoun Lim, Sung Whan Yoon
First submitted to arxiv on: 11 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 Federated learning (FL) has emerged as a prominent framework for training deep models on decentralized data while preserving privacy. To evaluate FL’s performance, researchers have established benchmarks that control data heterogeneity across clients. While existing benchmarks focus on simple classification tasks like MNIST and CIFAR, little attention has been paid to handling complex semantics with multiple labels. This paper proposes a novel approach to establishing an FL benchmark that handles semantic heterogeneity: it introduces two key steps – data clustering with semantics and data distributing via controllable semantic heterogeneity across clients. As proof of concept, the authors construct a federated Panoptic Scene Graph Generation (PSG) benchmark, demonstrating the effectiveness of existing PSG methods in an FL setting. The paper also showcases the performance gains achieved by robust federated learning algorithms in handling data heterogeneity. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research paper is about making computers learn from lots of different pieces of information without sharing the details. They want to make sure that the computer learning is fair and doesn’t have biases. The scientists are working on a new way to test how well this learning works by using real-life data with many different labels, not just simple ones like “cat” or “dog”. This will help them create more accurate and useful models for things like recognizing objects in pictures. They’re also showing that their new approach can make the computer learn better when dealing with mixed-up information. |
Keywords
» Artificial intelligence » Attention » Classification » Clustering » Federated learning » Semantics
