Distributed Multitask Learning
This addresses the problem of efficient distributed learning for related tasks in high-dimensional settings, representing an incremental improvement with specific gains.
The paper tackles distributed multi-task learning with high-dimensional linear predictors sharing a common support, proposing a communication-efficient estimator based on the debiased lasso that achieves performance comparable to optimal centralized methods.
We consider the problem of distributed multi-task learning, where each machine learns a separate, but related, task. Specifically, each machine learns a linear predictor in high-dimensional space,where all tasks share the same small support. We present a communication-efficient estimator based on the debiased lasso and show that it is comparable with the optimal centralized method.