Online Distributed Eigenspaces

Introduction

• Principal component analysis (PCA) is dimensionality reduction technique.
• We propose distributed eigenspaces to improve PCA performance.
• Our work is based on a recent paper that introduces the algorithm[1].
• We apply our implementation on CIFAR-10 dataset[2]

First point

Concepts of an online algorithm, The algorithm that is able to work with data that comes in a sequential manner.

Second point

The distributed algorithm, Kind of parallel algorithm that is designed to be run on hardware consisting of interconnected processors

Algorithm

References

1. D.Tarzanagh, M.Faradonbeh, G.Michailidis. Online Distributed Estimation of Principal Eigenspaces. 2019
2. Alex Krizhevsky. Learning Multiple Layers of Features from Tiny Images. 2009.