A topological insight into restricted Boltzmann machines

D.C. Mocanu, E. Mocanu, H.P. Nguyen, M. Gibescu, A. Liotta

Research output: Contribution to journalArticleAcademicpeer-review

26 Citations (Scopus)
183 Downloads (Pure)

Abstract

Restricted Boltzmann Machines (RBMs) and models derived from them have been successfully used as basic building blocks in deep artificial neural networks for automatic features extraction, unsupervised weights initialization, but also as density estimators. Thus, their generative and discriminative capabilities, but also their computational time are instrumental to a wide range of applications. Our main contribution is to look at RBMs from a topological perspective, bringing insights from network science. Firstly, here we show that RBMs and Gaussian RBMs (GRBMs) are bipartite graphs which naturally have a small-world topology. Secondly, we demonstrate both on synthetic and real-world datasets that by constraining RBMs and GRBMs to a scale-free topology (while still considering local neighborhoods and data distribution), we reduce the number of weights that need to be computed by a few orders of magnitude, at virtually no loss in generative performance. Thirdly, we show that, for a fixed number of weights, our proposed sparse models (which by design have a higher number of hidden neurons) achieve better generative capabilities than standard fully connected RBMs and GRBMs (which by design have a smaller number of hidden neurons), at no additional computational costs.
Original languageEnglish
Pages (from-to)243-270
Number of pages28
JournalMachine Learning
Volume104
Issue number2
DOIs
Publication statusPublished - Sep 2016

Fingerprint

Neurons
Topology
Feature extraction
Neural networks
Costs

Cite this

@article{c1346a3354ce4004afa0b003e978c887,
title = "A topological insight into restricted Boltzmann machines",
abstract = "Restricted Boltzmann Machines (RBMs) and models derived from them have been successfully used as basic building blocks in deep artificial neural networks for automatic features extraction, unsupervised weights initialization, but also as density estimators. Thus, their generative and discriminative capabilities, but also their computational time are instrumental to a wide range of applications. Our main contribution is to look at RBMs from a topological perspective, bringing insights from network science. Firstly, here we show that RBMs and Gaussian RBMs (GRBMs) are bipartite graphs which naturally have a small-world topology. Secondly, we demonstrate both on synthetic and real-world datasets that by constraining RBMs and GRBMs to a scale-free topology (while still considering local neighborhoods and data distribution), we reduce the number of weights that need to be computed by a few orders of magnitude, at virtually no loss in generative performance. Thirdly, we show that, for a fixed number of weights, our proposed sparse models (which by design have a higher number of hidden neurons) achieve better generative capabilities than standard fully connected RBMs and GRBMs (which by design have a smaller number of hidden neurons), at no additional computational costs.",
author = "D.C. Mocanu and E. Mocanu and H.P. Nguyen and M. Gibescu and A. Liotta",
year = "2016",
month = "9",
doi = "10.1007/s10994-016-5570-z",
language = "English",
volume = "104",
pages = "243--270",
journal = "Machine Learning",
issn = "0885-6125",
publisher = "Springer",
number = "2",

}

A topological insight into restricted Boltzmann machines. / Mocanu, D.C.; Mocanu, E.; Nguyen, H.P.; Gibescu, M.; Liotta, A.

In: Machine Learning, Vol. 104, No. 2, 09.2016, p. 243-270.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - A topological insight into restricted Boltzmann machines

AU - Mocanu, D.C.

AU - Mocanu, E.

AU - Nguyen, H.P.

AU - Gibescu, M.

AU - Liotta, A.

PY - 2016/9

Y1 - 2016/9

N2 - Restricted Boltzmann Machines (RBMs) and models derived from them have been successfully used as basic building blocks in deep artificial neural networks for automatic features extraction, unsupervised weights initialization, but also as density estimators. Thus, their generative and discriminative capabilities, but also their computational time are instrumental to a wide range of applications. Our main contribution is to look at RBMs from a topological perspective, bringing insights from network science. Firstly, here we show that RBMs and Gaussian RBMs (GRBMs) are bipartite graphs which naturally have a small-world topology. Secondly, we demonstrate both on synthetic and real-world datasets that by constraining RBMs and GRBMs to a scale-free topology (while still considering local neighborhoods and data distribution), we reduce the number of weights that need to be computed by a few orders of magnitude, at virtually no loss in generative performance. Thirdly, we show that, for a fixed number of weights, our proposed sparse models (which by design have a higher number of hidden neurons) achieve better generative capabilities than standard fully connected RBMs and GRBMs (which by design have a smaller number of hidden neurons), at no additional computational costs.

AB - Restricted Boltzmann Machines (RBMs) and models derived from them have been successfully used as basic building blocks in deep artificial neural networks for automatic features extraction, unsupervised weights initialization, but also as density estimators. Thus, their generative and discriminative capabilities, but also their computational time are instrumental to a wide range of applications. Our main contribution is to look at RBMs from a topological perspective, bringing insights from network science. Firstly, here we show that RBMs and Gaussian RBMs (GRBMs) are bipartite graphs which naturally have a small-world topology. Secondly, we demonstrate both on synthetic and real-world datasets that by constraining RBMs and GRBMs to a scale-free topology (while still considering local neighborhoods and data distribution), we reduce the number of weights that need to be computed by a few orders of magnitude, at virtually no loss in generative performance. Thirdly, we show that, for a fixed number of weights, our proposed sparse models (which by design have a higher number of hidden neurons) achieve better generative capabilities than standard fully connected RBMs and GRBMs (which by design have a smaller number of hidden neurons), at no additional computational costs.

U2 - 10.1007/s10994-016-5570-z

DO - 10.1007/s10994-016-5570-z

M3 - Article

VL - 104

SP - 243

EP - 270

JO - Machine Learning

JF - Machine Learning

SN - 0885-6125

IS - 2

ER -