Subspace projection approaches to classification and visualization of neural network-level encoding patterns

Remus Oşan, Liping Zhu, Shy Shoham, Joseph Zhuo Tsien

Research output: Contribution to journalArticle

12 Citations (Scopus)

Abstract

Recent advances in large-scale ensemble recordings allow monitoring of activity patterns of several hundreds of neurons in freely behaving animals. The emergence of such high-dimensional datasets poses challenges for the identification and analysis of dynamical network patterns. While several types of multivariate statistical methods have been used for integrating responses from multiple neurons, their effectiveness in pattern classification and predictive power has not been compared in a direct and systematic manner. Here we systematically employed a series of projection methods, such as Multiple Discriminant Analysis (MDA), Principal Components Analysis (PCA) and Artificial Neural Networks (ANN), and compared them with non-projection multivariate statistical methods such as Multivariate Gaussian Distributions (MGD). Our analyses of hippocampal data recorded during episodic memory events and cortical data simulated during face perception or arm movements illustrate how low-dimensional encoding subspaces can reveal the existence of network-level ensemble representations. We show how the use of regularization methods can prevent these statistical methods from over-fitting of training data sets when the trial numbers are much smaller than the number of recorded units. Moreover, we investigated the extent to which the computations implemented by the projection methods reflect the underlying hierarchical properties of the neural populations. Based on their ability to extract the essential features for pattern classification, we conclude that the typical performance ranking of these methods on under-sampled neural data of large dimension is MDA>PCA>ANN>MGD.

Original languageEnglish (US)
Article numbere404
JournalPloS one
Volume2
Issue number5
DOIs
StatePublished - May 2 2007

Fingerprint

neural networks
Statistical methods
Visualization
Gaussian distribution
Discriminant analysis
Neural networks
Principal component analysis
Neurons
Pattern recognition
statistical analysis
discriminant analysis
principal component analysis
neurons
Normal Distribution
Discriminant Analysis
Principal Component Analysis
Animals
methodology
Data storage equipment
Monitoring

ASJC Scopus subject areas

  • Biochemistry, Genetics and Molecular Biology(all)
  • Agricultural and Biological Sciences(all)

Cite this

Subspace projection approaches to classification and visualization of neural network-level encoding patterns. / Oşan, Remus; Zhu, Liping; Shoham, Shy; Tsien, Joseph Zhuo.

In: PloS one, Vol. 2, No. 5, e404, 02.05.2007.

Research output: Contribution to journalArticle

@article{b4051407dd924d5887ea62955ed66644,
title = "Subspace projection approaches to classification and visualization of neural network-level encoding patterns",
abstract = "Recent advances in large-scale ensemble recordings allow monitoring of activity patterns of several hundreds of neurons in freely behaving animals. The emergence of such high-dimensional datasets poses challenges for the identification and analysis of dynamical network patterns. While several types of multivariate statistical methods have been used for integrating responses from multiple neurons, their effectiveness in pattern classification and predictive power has not been compared in a direct and systematic manner. Here we systematically employed a series of projection methods, such as Multiple Discriminant Analysis (MDA), Principal Components Analysis (PCA) and Artificial Neural Networks (ANN), and compared them with non-projection multivariate statistical methods such as Multivariate Gaussian Distributions (MGD). Our analyses of hippocampal data recorded during episodic memory events and cortical data simulated during face perception or arm movements illustrate how low-dimensional encoding subspaces can reveal the existence of network-level ensemble representations. We show how the use of regularization methods can prevent these statistical methods from over-fitting of training data sets when the trial numbers are much smaller than the number of recorded units. Moreover, we investigated the extent to which the computations implemented by the projection methods reflect the underlying hierarchical properties of the neural populations. Based on their ability to extract the essential features for pattern classification, we conclude that the typical performance ranking of these methods on under-sampled neural data of large dimension is MDA>PCA>ANN>MGD.",
author = "Remus Oşan and Liping Zhu and Shy Shoham and Tsien, {Joseph Zhuo}",
year = "2007",
month = "5",
day = "2",
doi = "10.1371/journal.pone.0000404",
language = "English (US)",
volume = "2",
journal = "PLoS One",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "5",

}

TY - JOUR

T1 - Subspace projection approaches to classification and visualization of neural network-level encoding patterns

AU - Oşan, Remus

AU - Zhu, Liping

AU - Shoham, Shy

AU - Tsien, Joseph Zhuo

PY - 2007/5/2

Y1 - 2007/5/2

N2 - Recent advances in large-scale ensemble recordings allow monitoring of activity patterns of several hundreds of neurons in freely behaving animals. The emergence of such high-dimensional datasets poses challenges for the identification and analysis of dynamical network patterns. While several types of multivariate statistical methods have been used for integrating responses from multiple neurons, their effectiveness in pattern classification and predictive power has not been compared in a direct and systematic manner. Here we systematically employed a series of projection methods, such as Multiple Discriminant Analysis (MDA), Principal Components Analysis (PCA) and Artificial Neural Networks (ANN), and compared them with non-projection multivariate statistical methods such as Multivariate Gaussian Distributions (MGD). Our analyses of hippocampal data recorded during episodic memory events and cortical data simulated during face perception or arm movements illustrate how low-dimensional encoding subspaces can reveal the existence of network-level ensemble representations. We show how the use of regularization methods can prevent these statistical methods from over-fitting of training data sets when the trial numbers are much smaller than the number of recorded units. Moreover, we investigated the extent to which the computations implemented by the projection methods reflect the underlying hierarchical properties of the neural populations. Based on their ability to extract the essential features for pattern classification, we conclude that the typical performance ranking of these methods on under-sampled neural data of large dimension is MDA>PCA>ANN>MGD.

AB - Recent advances in large-scale ensemble recordings allow monitoring of activity patterns of several hundreds of neurons in freely behaving animals. The emergence of such high-dimensional datasets poses challenges for the identification and analysis of dynamical network patterns. While several types of multivariate statistical methods have been used for integrating responses from multiple neurons, their effectiveness in pattern classification and predictive power has not been compared in a direct and systematic manner. Here we systematically employed a series of projection methods, such as Multiple Discriminant Analysis (MDA), Principal Components Analysis (PCA) and Artificial Neural Networks (ANN), and compared them with non-projection multivariate statistical methods such as Multivariate Gaussian Distributions (MGD). Our analyses of hippocampal data recorded during episodic memory events and cortical data simulated during face perception or arm movements illustrate how low-dimensional encoding subspaces can reveal the existence of network-level ensemble representations. We show how the use of regularization methods can prevent these statistical methods from over-fitting of training data sets when the trial numbers are much smaller than the number of recorded units. Moreover, we investigated the extent to which the computations implemented by the projection methods reflect the underlying hierarchical properties of the neural populations. Based on their ability to extract the essential features for pattern classification, we conclude that the typical performance ranking of these methods on under-sampled neural data of large dimension is MDA>PCA>ANN>MGD.

UR - http://www.scopus.com/inward/record.url?scp=55849091323&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=55849091323&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0000404

DO - 10.1371/journal.pone.0000404

M3 - Article

C2 - 17476326

AN - SCOPUS:55849091323

VL - 2

JO - PLoS One

JF - PLoS One

SN - 1932-6203

IS - 5

M1 - e404

ER -