Unsupervised Deep Learning for Subspace Clustering

Date

2017-12-14

Author

SEKMEN, ali
Koku, Ahmet Buğra
PARLAKTUNA, Mustafa
ABDULMALEK, Ayad
VANAMALA, Nagendrababu

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

192
views

0
downloads

This paper presents a novel technique for the segmentation of data W = [w(1) . . . w(N)] subset of R-D drawn from a union U = boolean OR(M)(i=1) S-i of subspaces {S-i}(i=1)(M). First, an existing subspace segmentation algorithm is used to perform an initial data clustering {C-i}(i=1)(M), where C-i = {w(i1) . . . w(ik)} subset of W is the set of data from the ith cluster. Then, a local subspace LSi is matched for each C-i and the distance d(ij) between LSi and each point w(ij) is an element of C-i is computed. A data-driven threshold eta is computed and the data points (in C-i) whose distances to LSi are larger than eta are eliminated since they are considered as outliers or erroneously clustered data points in C-i. The remaining data points (C) over tilde (i) subset of C-i are considered to be coming from the same subspace with high confidence. Then, {(C) over tilde (i)}(i=1)(M) are used in unsupervised way to train a convolution neural network to obtain a deep learning model, which is in turn used to re-cluster W. The system has been successfully implemented using the MNIST dataset and it improved the segmentation accuracy of a particular algorithm (EnSC-ORGEN) from 93.79% to 96.52%.

Subject Keywords

Deep learning, Convolution neural networks, Data clustering, Subspace segmentation, Union of subspaces

URI

https://hdl.handle.net/11511/55782

Collections

Department of Mechanical Engineering, Conference / Seminar

Suggestions

OpenMETU
Core

Similarity matrix framework for data from union of subspaces Aldroubi, Akram; Sekmen, Ali; Koku, Ahmet Buğra; Cakmak, Ahmet Faruk (2018-09-01) This paper presents a framework for finding similarity matrices for the segmentation of data W = [w(1)...w(N)] subset of R-D drawn from a union U = boolean OR(M)(i=1) S-i, of independent subspaces {S-i}(i=1)(M), of dimensions {d(i)}(i=1)(M). It is shown that any factorization of W = BP, where columns of B form a basis for data W and they also come from U, can be used to produce a similarity matrix Xi w. In other words, Xi w(i, j) not equal 0, when the columns w(i) and w(j) of W come from the same subspace, ...
Unsupervised segmentation of hyperspectral images using modified phase correlation Ertuerk, Alp; Ertuerk, Sarp (2006-10-01) This letter presents hyperspectral image segmentation based on the phase-correlation measure of subsampled hyperspectral data, which is referred to as modified phase correlation. The hyperspectral spectrum of each pixel is initially subsampled to gain, robustness against noise and spatial variability, and phase correlation is applied to determine spectral similarity. Similar and dissimilar pixels are decided according to the peak value of the phase correlation result to determine pixels that fall into the s...
Deep Spectral Convolution Network for Hyperspectral Unmixing Akar, Gözde (2018-10-10) In this paper, we propose a novel hyperspectral unmixing technique based on deep spectral convolution networks (DSCN). Particularly, three important contributions are presented throughout this paper. First, fully-connected linear operation is replaced with spectral convolutions to extract local spectral characteristics from hyperspectral signatures with a deeper network architecture. Second, instead of batch normalization, we propose a spectral normalization layer which improves the selectivity of filters b...
A Trie-structured Bayesian Model for Unsupervised Morphological Segmentation Kurfalı, Murathan; Ustun, Ahmet; CAN BUĞLALILAR, BURCU (2017-04-23) In this paper, we introduce a trie-structured Bayesian model for unsupervised morphological segmentation. We adopt prior information from different sources in the model. We use neural word embeddings to discover words that are morphologically derived from each other and thereby that are semantically similar. We use letter successor variety counts obtained from tries that are built by neural word embeddings. Our results show that using different information sources such as neural word embeddings and letter s...
Skeleton Decomposition Analysis for Subspace Clustering Sekmen, Ali; Aldroubi, Akram; Koku, Ahmet Buğra (2016-12-08) This paper provides a comprehensive analysis of skeleton decomposition used for segmentation of data W = [w(1) center dot center dot center dot w(N)] subset of R-D drawn from a union u = U-i=1(M) S-i of linearly independent subspaces {Si}(M)(i=1) of dimensionsof {di}(M)(i=1). Our previous work developed a generalized theoretical framework for computing similarity matrices by matrix factorization. Skeleton decomposition is a special case of this general theory. First, a square sub-matrix A is an element of R...

Citation Formats

a. SEKMEN, A. B. Koku, M. PARLAKTUNA, A. ABDULMALEK, and N. VANAMALA, “Unsupervised Deep Learning for Subspace Clustering,” 2017, Accessed: 00, 2020. [Online]. Available: https://hdl.handle.net/11511/55782.