Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-Sets
Dimension reduction is often used for several procedures of analysis of high dimensional biomedical data-sets such as classification or outlier detection. To improve the performance of such data-mining steps, preserving both distance information and local topology among data-points could be more use...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-08-01
|
Series: | Algorithms |
Subjects: | |
Online Access: | https://www.mdpi.com/1999-4893/13/8/192 |
id |
doaj-3abc8594e80c4179869b906ef2d54d21 |
---|---|
record_format |
Article |
spelling |
doaj-3abc8594e80c4179869b906ef2d54d212020-11-25T03:37:11ZengMDPI AGAlgorithms1999-48932020-08-011319219210.3390/a13080192Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-SetsKaraj Khosla0Indra Prakash Jha1Ajit Kumar2Vibhor Kumar3Department of Computer Science, Guru Tegh Bahadur Institute of Technology, New Delhi 110064, IndiaDepartment of Computational Biology, IIIT Delhi, Okhla Phase-3, New Delhi 110020, IndiaAdobe, Block-A, Sector-321, Noida 201304, IndiaDepartment of Computational Biology, IIIT Delhi, Okhla Phase-3, New Delhi 110020, IndiaDimension reduction is often used for several procedures of analysis of high dimensional biomedical data-sets such as classification or outlier detection. To improve the performance of such data-mining steps, preserving both distance information and local topology among data-points could be more useful than giving priority to visualization in low dimension. Therefore, we introduce topology-preserving distance scaling (TPDS) to augment a dimension reduction method meant to reproduce distance information in a higher dimension. Our approach involves distance inflation to preserve local topology to avoid collapse during distance preservation-based optimization. Applying TPDS on diverse biomedical data-sets revealed that besides providing better visualization than typical distance preserving methods, TPDS leads to better classification of data points in reduced dimension. For data-sets with outliers, the approach of TPDS also proves to be useful, even for purely distance-preserving method for achieving better convergence.https://www.mdpi.com/1999-4893/13/8/192dimension reductiondistance preservinglocal topologymultidimensional scaling (MDS) |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Karaj Khosla Indra Prakash Jha Ajit Kumar Vibhor Kumar |
spellingShingle |
Karaj Khosla Indra Prakash Jha Ajit Kumar Vibhor Kumar Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-Sets Algorithms dimension reduction distance preserving local topology multidimensional scaling (MDS) |
author_facet |
Karaj Khosla Indra Prakash Jha Ajit Kumar Vibhor Kumar |
author_sort |
Karaj Khosla |
title |
Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-Sets |
title_short |
Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-Sets |
title_full |
Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-Sets |
title_fullStr |
Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-Sets |
title_full_unstemmed |
Local-Topology-Based Scaling for Distance Preserving Dimension Reduction Method to Improve Classification of Biomedical Data-Sets |
title_sort |
local-topology-based scaling for distance preserving dimension reduction method to improve classification of biomedical data-sets |
publisher |
MDPI AG |
series |
Algorithms |
issn |
1999-4893 |
publishDate |
2020-08-01 |
description |
Dimension reduction is often used for several procedures of analysis of high dimensional biomedical data-sets such as classification or outlier detection. To improve the performance of such data-mining steps, preserving both distance information and local topology among data-points could be more useful than giving priority to visualization in low dimension. Therefore, we introduce topology-preserving distance scaling (TPDS) to augment a dimension reduction method meant to reproduce distance information in a higher dimension. Our approach involves distance inflation to preserve local topology to avoid collapse during distance preservation-based optimization. Applying TPDS on diverse biomedical data-sets revealed that besides providing better visualization than typical distance preserving methods, TPDS leads to better classification of data points in reduced dimension. For data-sets with outliers, the approach of TPDS also proves to be useful, even for purely distance-preserving method for achieving better convergence. |
topic |
dimension reduction distance preserving local topology multidimensional scaling (MDS) |
url |
https://www.mdpi.com/1999-4893/13/8/192 |
work_keys_str_mv |
AT karajkhosla localtopologybasedscalingfordistancepreservingdimensionreductionmethodtoimproveclassificationofbiomedicaldatasets AT indraprakashjha localtopologybasedscalingfordistancepreservingdimensionreductionmethodtoimproveclassificationofbiomedicaldatasets AT ajitkumar localtopologybasedscalingfordistancepreservingdimensionreductionmethodtoimproveclassificationofbiomedicaldatasets AT vibhorkumar localtopologybasedscalingfordistancepreservingdimensionreductionmethodtoimproveclassificationofbiomedicaldatasets |
_version_ |
1724546629802917888 |