Communication-Efficient Agnostic Federated Averaging
This work addresses bias in federated learning for large-scale, cross-device applications, such as language modeling on millions of devices, but it is incremental as it builds on prior domain-agnostic methods.
The paper tackles the problem of bias in federated learning by proposing AgnosticFedAvg, a communication-efficient algorithm for the cross-device setting with many clients, and demonstrates its effectiveness through large-scale language modeling tasks, achieving positive results in simulations and live experiments involving millions of user devices.
In distributed learning settings such as federated learning, the training algorithm can be potentially biased towards different clients. Mohri et al. (2019) proposed a domain-agnostic learning algorithm, where the model is optimized for any target distribution formed by a mixture of the client distributions in order to overcome this bias. They further proposed an algorithm for the cross-silo federated learning setting, where the number of clients is small. We consider this problem in the cross-device setting, where the number of clients is much larger. We propose a communication-efficient distributed algorithm called Agnostic Federated Averaging (or AgnosticFedAvg) to minimize the domain-agnostic objective proposed in Mohri et al. (2019), which is amenable to other private mechanisms such as secure aggregation. We highlight two types of naturally occurring domains in federated learning and argue that AgnosticFedAvg performs well on both. To demonstrate the practical effectiveness of AgnosticFedAvg, we report positive results for large-scale language modeling tasks in both simulation and live experiments, where the latter involves training language models for Spanish virtual keyboard for millions of user devices.