Fast and Robust Subspace Clustering Using Random Projections

03/30/2018
by   Guangcan Liu, et al.
0

Over the past several decades, subspace clustering has been receiving increasing interest and continuous progress. However, due to the lack of scalability and/or robustness, existing methods still have difficulty in dealing with the data that possesses simultaneously three characteristics: high-dimensional, massive and grossly corrupted. To tackle the scalability and robustness issues simultaneously, in this paper we suggest to consider a problem called compressive robust subspace clustering, which is to perform robust subspace clustering with the compressed data, and which is generated by projecting the original high-dimensional data onto a lower-dimensional subspace chosen at random. Given these random projections, the proposed method, row space pursuit (RSP), recovers not only the authentic row space, which provably leads to correct clustering results under certain conditions, but also the gross errors possibly existing in data. The compressive nature of the random projections gives our RSP high computational and storage efficiency, and the recovery property enables the ability for RSP to deal with the grossly corrupted data. Extensive experiments on high-dimensional and/or large-scale datasets show that RSP can maintain comparable accuracies to to prevalent methods with significant reductions in the computational time.

READ FULL TEXT

page 6

page 7

page 10

research
04/27/2014

Subspace clustering of dimensionality-reduced data

Subspace clustering refers to the problem of clustering unlabeled high-d...
research
05/23/2019

Johnson-Lindenstrauss Property Implies Subspace Restricted Isometry Property

Dimensionality reduction is a popular approach to tackle high-dimensiona...
research
03/05/2020

Simultaneous robust subspace recovery and semi-stability of quiver representations

We consider the problem of simultaneously finding lower-dimensional subs...
research
09/24/2019

High-dimensional clustering via Random Projections

In this work, we address the unsupervised classification issue by exploi...
research
12/03/2019

A Fast deflation Method for Sparse Principal Component Analysis via Subspace Projections

Deflation method is an iterative technique that searches the sparse load...
research
07/22/2017

Sketched Subspace Clustering

The immense amount of daily generated and communicated data presents uni...
research
09/16/2015

Projection Bank: From High-dimensional Data to Medium-length Binary Codes

Recently, very high-dimensional feature representations, e.g., Fisher Ve...

Please sign up or login with your details

Forgot password? Click here to reset