University of Illinois Urbana-Champaign

Application of CNN auto-encoder in Spatial Dimensionality Reduction of Mass Spectrometry Imaging Data

Liu, Liyu

This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.

Permalink

https://hdl.handle.net/2142/109145

Description

Title
Application of CNN auto-encoder in Spatial Dimensionality Reduction of Mass Spectrometry Imaging Data
Author(s)
Liu, Liyu
Contributor(s)
Ochoa, Idoia
Issue Date
2020-12
Keyword(s)
  • CNN Auto-Encoder
  • Spatial Dimensionality Reduction
  • Mass Spectrometry Imaging Data
Abstract
Methods to reduce data spatial dimension so as to reduce the computational cost while preserving the original spatial information are an important research field in machine learning. Methods such as PCA (Principal Component Analysis) and NMF (non-negative matrix factorization) are widely used for dimension reduction. There are also works on using auto-encoder based on fully connected layers for dimension reduction. However, fully connected layers have extremely high computational cost and are therefore time consuming. In this work, we propose a CNN layers based auto-encoder to reduce the dimension of data input. We applied our network on Mouse Urinary Bladder Data-set with fixed dimensions of 5180* 38440. We compared the performance regarding execution time, MSE and spatial variance with those of PCA, NMF and FC auto-encoder. Our result shows that although we sacrificed MSE accuracy compared with PCA, NMF and FC auto-encoder, we preserved the most spatial information compared with raw input. Also, our CNN layers based auto-encoder increased the computational speed by almost 4x compared to that of FC auto-encoder. Our result promotes the potential of using the reduced feature for further calculation.
Type of Resource
text
Language
en
Permalink
http://hdl.handle.net/2142/109145

Owning Collections

Senior Theses - Electrical and Computer Engineering PRIMARY
The best of ECE undergraduate research