Abstract
The rapid development in data collecting devices and computation platforms produces an emerging number of agents, each equipped with a unique data modality over a particular population of subjects. While an agent’s predictive performance may be enhanced by transmitting others’ data to it, this is often unrealistic due to intractable transmission costs and security concerns. In this paper, we propose a method named ASCII for an agent to improve its classification performance through assistance from other agents, without sharing proprietary data and model information. The main idea is to iteratively interchange an ignorance value between 0 and 1 for each collated sample among agents, where the value represents the urgency of further assistance needed. The method is naturally suitable for privacy-aware, transmission-economical, and decentralized learning scenarios. The method is also general as it allows the agents to use arbitrary classifiers such as logistic regression, ensemble tree, and neural network, and they may be heterogeneous among agents. We demonstrate the proposed method with extensive experimental studies.
A Generalization for Classification
In this paper, we proposed a general method for an agent to improve its classification performance by iteratively interchanging ignorance scores with other agents. Our method is naturally suitable for autonomous learning scenarios where private raw data cannot be shared. Moreover, the proposed method allows agents to use private local models or algorithms, which is appealing in many application domains. Some future directions are summarized as follows. First, our work addressed classification, and we believe that similar techniques can be emulated to study regression problems. Second, from various experiments, we observed that a single agent often achieves nearoracle performance by interchanging with other agents in random orders. This motivates the problem to study the most efficient order of interchanging information to attain the optimum. Another open problem is to study how asynchronous interchange, meaning different orders at two rounds, will influence the learning efficiency.
References
- Zhou, Jiaying, Xun Xian, Na Li, and Jie Ding. “Assisted learning: cooperative AI with autonomy.” In ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 3130-3134. IEEE, 2021. [DOC]