Linear Speedup in Personalized Collaborative Learning. (arXiv:2111.05968v4 [cs.LG] UPDATED)

Web: http://arxiv.org/abs/2111.05968

Collaborative training can improve the accuracy of a model for a user by
trading off the model's bias (introduced by using data from other users who are
potentially different) against its variance (due to the limited amount of data
on any single user). In this work, we formalize the personalized collaborative
learning problem as a stochastic optimization of a task 0 while giving access
to N related but different tasks 1,..., N. We provide convergence guarantees
for two algorithms in …

arxiv collaborative learning lg linear personalized