Summary of Tabvfl: Improving Latent Representation in Vertical Federated Learning, by Mohamed Rashad et al.
TabVFL: Improving Latent Representation in Vertical Federated Learning
by Mohamed Rashad, Zilong Zhao, Jeremie Decouchant, Lydia Y. Chen
First submitted to arxiv on: 27 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Distributed, Parallel, and Cluster Computing (cs.DC)
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 TabNet, a state-of-the-art neural network model designed for tabular data, utilizes an autoencoder architecture for training. In Vertical Federated Learning (VFL), multiple parties train a model collaboratively on vertically partitioned data while maintaining data privacy. The existing design of training autoencoders in VFL could potentially break correlations between feature data of participating parties. Traditional autoencoders are not designed for tabular data, which is ubiquitous in VFL settings. Furthermore, the impact of client failures during training on model robustness is under-researched in the VFL scene. This paper proposes TabVFL, a distributed framework that improves latent representation learning using joint features of participants. The framework preserves privacy by mitigating potential data leakage, conserves feature correlations by learning one latent representation vector, and provides enhanced robustness against client failures during training. Extensive experiments on five classification datasets show that TabVFL outperforms prior work design with 26.12% improvement on f1-score. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary TabNet is a type of neural network model used for tabular data. In Vertical Federated Learning, different parties share data and train a model together without sharing their individual data. The current way to do this could cause problems because it doesn’t take into account the relationships between features. Traditional autoencoders aren’t designed for this kind of data, which is common in VFL. This paper proposes a new way to do this called TabVFL. It preserves privacy by hiding sensitive information, keeps track of feature relationships by using one type of representation, and makes the model more robust against problems during training. |
Keywords
» Artificial intelligence » Autoencoder » Classification » F1 score » Federated learning » Neural network » Representation learning
