A Testbed for Neural-Network Models Capable of Integrating Information
in Time
S. Zappacosta, S. Nolfi, and G. Baldassarre. Anticipatory Behavior in Adaptive Learning Systems: From Brains
to Individual and Social Behavior, Springer-Verlag, (2007)
Abstract
This paper presents a set of techniques that allow generating
a class of testbeds that can be used to test recurrent neural networks
capabilities of integrating information in time. More in particular,
the
testbeds allow evaluating the capability of such models, and possibly
other architectures and algorithms, of (a) categorizing different
time series,
(b) anticipating future signal levels on the basis of past ones, and
(c)
functioning robustly with respect to noise and other systematic random
variations of the temporal and spatial properties of the input time
series.
The paper also presents a number of analysis tools that can be used
to
understand the functioning and organization of the dynamical internal
representations that recurrent neural networks develop to acquire
the
aforementioned capabilities, for example to understand how they capture
time regularities such as periodicity, repetitions, spikes, numbers,
levels and rates of change of input signals. The utility of the proposed
testbeds is illustrated by testing and studying the capacity of Elman
neural networks to predict and categorize different signals in two
simple
tasks.
%0 Book Section
%1 Zappacosta:2007a
%A Zappacosta, Stefano
%A Nolfi, Stefano
%A Baldassarre, Gianluca
%B Anticipatory Behavior in Adaptive Learning Systems: From Brains
to Individual and Social Behavior
%D 2007
%E Butz, Martin V.
%E Sigaud, Olivier
%E Pezzulo, Giovanni
%E Baldassarre, Gianluca
%I Springer-Verlag
%K Analysis Analysis, Attractors Information Integration, Internal Networks,Dynamical Neural Recurrent Representations, Series, Signal Systems, Testbed, Time Waves, of
%T A Testbed for Neural-Network Models Capable of Integrating Information
in Time
%X This paper presents a set of techniques that allow generating
a class of testbeds that can be used to test recurrent neural networks
capabilities of integrating information in time. More in particular,
the
testbeds allow evaluating the capability of such models, and possibly
other architectures and algorithms, of (a) categorizing different
time series,
(b) anticipating future signal levels on the basis of past ones, and
(c)
functioning robustly with respect to noise and other systematic random
variations of the temporal and spatial properties of the input time
series.
The paper also presents a number of analysis tools that can be used
to
understand the functioning and organization of the dynamical internal
representations that recurrent neural networks develop to acquire
the
aforementioned capabilities, for example to understand how they capture
time regularities such as periodicity, repetitions, spikes, numbers,
levels and rates of change of input signals. The utility of the proposed
testbeds is illustrated by testing and studying the capacity of Elman
neural networks to predict and categorize different signals in two
simple
tasks.
@incollection{Zappacosta:2007a,
abstract = {This paper presents a set of techniques that allow generating
a class of testbeds that can be used to test recurrent neural networks
capabilities of integrating information in time. More in particular,
the
testbeds allow evaluating the capability of such models, and possibly
other architectures and algorithms, of (a) categorizing different
time series,
(b) anticipating future signal levels on the basis of past ones, and
(c)
functioning robustly with respect to noise and other systematic random
variations of the temporal and spatial properties of the input time
series.
The paper also presents a number of analysis tools that can be used
to
understand the functioning and organization of the dynamical internal
representations that recurrent neural networks develop to acquire
the
aforementioned capabilities, for example to understand how they capture
time regularities such as periodicity, repetitions, spikes, numbers,
levels and rates of change of input signals. The utility of the proposed
testbeds is illustrated by testing and studying the capacity of Elman
neural networks to predict and categorize different signals in two
simple
tasks.},
added-at = {2009-06-26T15:25:19.000+0200},
author = {Zappacosta, Stefano and Nolfi, Stefano and Baldassarre, Gianluca},
biburl = {https://www.bibsonomy.org/bibtex/263439db958be83fb2e5c94ea519af48b/butz},
booktitle = {Anticipatory Behavior in Adaptive Learning Systems: {F}rom Brains
to Individual and Social Behavior},
description = {diverse cognitive systems bib},
editor = {Butz, Martin V. and Sigaud, Olivier and Pezzulo, Giovanni and Baldassarre, Gianluca},
interhash = {46a51538e7b7e8980c82fbaf1faf3cc4},
intrahash = {63439db958be83fb2e5c94ea519af48b},
keywords = {Analysis Analysis, Attractors Information Integration, Internal Networks,Dynamical Neural Recurrent Representations, Series, Signal Systems, Testbed, Time Waves, of},
owner = {butz},
publisher = {Springer-{V}erlag},
timestamp = {2009-06-26T15:26:02.000+0200},
title = {A Testbed for Neural-Network Models Capable of Integrating Information
in Time},
year = 2007
}