Communication-Efficient Distributed Online Learning with Kernels
This work addresses communication bottlenecks in real-time distributed systems, offering an incremental improvement for kernel-based online learning.
The paper tackles the inefficiency of communicating support vector expansions in distributed online learning by extending a prior protocol to kernelized learners, enabling model compression and introducing a novel communication criterion bounded by loss.
We propose an efficient distributed online learning protocol for low-latency real-time services. It extends a previously presented protocol to kernelized online learners that represent their models by a support vector expansion. While such learners often achieve higher predictive performance than their linear counterparts, communicating the support vector expansions becomes inefficient for large numbers of support vectors. The proposed extension allows for a larger class of online learning algorithms---including those alleviating the problem above through model compression. In addition, we characterize the quality of the proposed protocol by introducing a novel criterion that requires the communication to be bounded by the loss suffered.