Adaptive Classification of Temporal Signals in Fixed-Weight Recurrent Neural Networks: An Existence Proof

DOI10.1162/NECO.2008.05-07-527MaRDI QIDQ3527192zbMATH OpenOpenAlexFDO

Authors Danil Prokhorov, C. van Leeuwen, I. Tyukin

Publication date 25 September 2008

Published in Neural Computation (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/0705.3370

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05)

Abstract: We address the important theoretical question why a recurrent neural network with fixed weights can adaptively classify time-varied signals in the presence of additive noise and parametric perturbations. We provide a mathematical proof assuming that unknown parameters are allowed to enter the signal nonlinearly and the noise amplitude is sufficiently small.

Recommendations

• Adaptive Capability of Recurrent Neural Networks with Fixed Weights for Series-Parallel System Identification
• Learning periodic signals with recurrent neural networks
• Recurrent neural networks for time series classification
• Local stability of recurrent networks with time-varying weights and inputs
• A recurrent neural network with ever changing synapses

Cites work

• Absolute stability of global pattern formation and parallel memory storage by competitive neural networks
• Adaptation and Parameter Estimation in Systems With Unstable Target Dynamics and Nonlinear Parametrization
• Adaptation and regulation with signal detection implies internal model
• Approximation by superpositions of a sigmoidal function
• Chaotic itinerancy
• Every good regulator of a system must be a model of that system †
• Neural networks and physical systems with emergent collective computational abilities
• New Conditions on Global Stability of Cohen-Grossberg Neural Networks
• Nonuniform Small-Gain Theorems for Systems with Unstable Invariant Sets
• On the concept of attractor
• Parameter convergence in nonlinearly parameterized systems
• Qualitative analysis of neural networks
• Universal Approximation of Multiple Nonlinear Operators by Neural Networks

Cited in

(2)

• Adaptive Capability of Recurrent Neural Networks with Fixed Weights for Series-Parallel System Identification
• Sampled-data control for Markovian switching neural networks with output quantization and packet dropouts

This page was built for publication: Adaptive Classification of Temporal Signals in Fixed-Weight Recurrent Neural Networks: An Existence Proof

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3527192)