LOCO: Distributing Ridge Regression with Random Projections
This addresses efficient distributed computation for large-scale regression problems, though it appears incremental as an optimization of existing distributed ridge regression approaches.
The authors tackled large-scale ridge regression by proposing LOCO, a distributed algorithm that preserves feature dependencies using structured random projections. They showed LOCO produces solutions close to exact ridge regression and achieved significant speedups compared to state-of-the-art distributed methods in experiments including climate prediction.
We propose LOCO, an algorithm for large-scale ridge regression which distributes the features across workers on a cluster. Important dependencies between variables are preserved using structured random projections which are cheap to compute and must only be communicated once. We show that LOCO obtains a solution which is close to the exact ridge regression solution in the fixed design setting. We verify this experimentally in a simulation study as well as an application to climate prediction. Furthermore, we show that LOCO achieves significant speedups compared with a state-of-the-art distributed algorithm on a large-scale regression problem.