conftrace_

2018 ICML ICML 2018

Do Outliers Ruin Collaboration?

Abstract

We consider the problem of learning a binary classifier from $n$ different data sources, among which at most an $\eta$ fraction are adversarial. The overhead is defined as the ratio between the sample complexity of learning in this setting and that of learning the same hypothesis class on a single data distribution. We present an algorithm that achieves an $O(\eta n + \ln n)$ overhead, which is proved to be worst-case optimal. We also discuss the potential challenges to the design of a computationally efficient learning algorithm with a small overhead.

❓ The Questioner

📈 Trend Setter - Robust Learning

🐣 Hot Topic Early Bird - collaborative learning

Authors

Mingda Qiao

Topics

Artificial Intelligence > Learning Paradigms > Federated Learning Machine Learning > Learning Types > Adversarial Learning Machine Learning > Optimization & Theory > Learning Theory Machine Learning > Learning Types > Robust Learning Machine Learning > Learning Types > Multi-Source Learning

Keywords

federated learning sample complexity distributed learning collaborative learning robust learning binary classifier adversarial noise adversarial outlier

More from ICML 2018

All ICML 2018 papers →

JointGAN: Multi-Domain Joint Distribution Learning with Generative Adversarial Nets

Selecting Representative Examples for Program Synthesis

Generalized Earley Parser: Bridging Symbolic Grammars and Sequence Data for Future Prediction

Gradually Updated Neural Networks for Large-Scale Image Recognition

DCFNet: Deep Neural Network with Decomposed Convolutional Filters

Non-convex Conditional Gradient Sliding