Assisted Learning for Organizations with Limited Imbalanced Data
Cheng Chen, Jiaying Zhou, Jie Ding, Yi Zhou
TL;DR
The paper tackles learning when an organization has limited, imbalanced data and cannot share raw data with external providers. It proposes a horizontal-splitting assisted-learning framework and introduces two algorithms, AssistDeep for deep learning and AssistPG for reinforcement learning, which enable near-oracle performance with only a few interaction rounds by exchanging model trajectories and losses rather than data. It provides convergence guarantees for AssistDeep under standard smoothness assumptions and demonstrates strong empirical results on CIFAR-10, SVHN, CartPole, and LunarLander, showing substantial gains over training on the learner’s data alone and robustness to data imbalance and limited communications. The framework offers a practical pathway for organizations to improve ML performance while preserving data privacy, with potential extensions to meta-learning and multi-agent configurations in future work.
Abstract
In the era of big data, many big organizations are integrating machine learning into their work pipelines to facilitate data analysis. However, the performance of their trained models is often restricted by limited and imbalanced data available to them. In this work, we develop an assisted learning framework for assisting organizations to improve their learning performance. The organizations have sufficient computation resources but are subject to stringent data-sharing and collaboration policies. Their limited imbalanced data often cause biased inference and sub-optimal decision-making. In assisted learning, an organizational learner purchases assistance service from an external service provider and aims to enhance its model performance within only a few assistance rounds. We develop effective stochastic training algorithms for both assisted deep learning and assisted reinforcement learning. Different from existing distributed algorithms that need to frequently transmit gradients or models, our framework allows the learner to only occasionally share information with the service provider, but still obtain a model that achieves near-oracle performance as if all the data were centralized.