Summary of Fedne: Surrogate-assisted Federated Neighbor Embedding For Dimensionality Reduction, by Ziwei Li et al.
FedNE: Surrogate-Assisted Federated Neighbor Embedding for Dimensionality Reduction
by Ziwei Li, Xiaoqi Wang, Hong-You Chen, Han-Wei Shen, Wei-Lun Chao
First submitted to arxiv on: 17 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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) is a collaborative model training approach that enables distributed participants to share knowledge without exchanging local data. While FL has numerous applications, visualizing complex high-dimensional data remains a significant challenge. Neighbor embedding (NE) is crucial for this task, but collaboratively learning a joint NE model is difficult due to the lack of effective visualization algorithms and the need for computing loss functions among pairs of data. This paper introduces FedNE, an innovative approach that combines the FedAvg framework with contrastive NE technique, without requiring shareable data. To address the issue of inter-client repulsion, a surrogate loss function is developed to enable each client to learn and share information with others. Additionally, a data-mixing strategy is proposed to relax problems related to invisible neighbors and false neighbors constructed by local kNN graphs. The paper conducts comprehensive experiments on both synthetic and real-world datasets, demonstrating that FedNE can effectively preserve neighborhood data structures and enhance alignment in the global embedding space compared to several baseline methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine a way for different groups of people to work together on a project without sharing their individual information. This is called federated learning (FL). FL is important because it helps us solve big problems, like recognizing pictures or understanding languages. However, there’s a challenge: how do we show what all these individual pieces of information look like when they’re combined? Neighbor embedding (NE) is a technique that helps with this. But, if many people are working together on this task, it gets harder to figure out how to make everything align. This paper introduces FedNE, a new way to solve this problem using FL and NE techniques. It’s like creating a map that shows how all the individual pieces fit together. The results show that FedNE is effective in showing these connections compared to other methods. |
Keywords
» Artificial intelligence » Alignment » Embedding » Embedding space » Federated learning » Loss function
