A biological gradient descent for prediction through a combination of STDP and homeostatic plasticity

DOI10.1162/NECO_A_00512MaRDI QIDQ5378278zbMATH OpenWikidataFDO

Authors Gilles Wainrib, Mathieu N. Galtier

Publication date 12 June 2019

Published in Neural Computation (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1206.4812

zbMATH Keywords

homeostatic plasticitypredictionrecurrent neural networksspike-timing-dependent plasticity

Mathematics Subject Classification ID

Learning and adaptive systems in artificial intelligence (68T05) Neural networks for/in biological studies, artificial life and related topics (92B20)

Abstract: Identifying, formalizing and combining biological mechanisms which implement known brain functions, such as prediction, is a main aspect of current research in theoretical neuroscience. In this letter, the mechanisms of Spike Timing Dependent Plasticity (STDP) and homeostatic plasticity, combined in an original mathematical formalism, are shown to shape recurrent neural networks into predictors. Following a rigorous mathematical treatment, we prove that they implement the online gradient descent of a distance between the network activity and its stimuli. The convergence to an equilibrium, where the network can spontaneously reproduce or predict its stimuli, does not suffer from bifurcation issues usually encountered in learning in recurrent neural networks.

Recommendations

• Evaluating the extent to which homeostatic plasticity learns to compute prediction errors in unstructured neuronal networks
• Adaptive Synchronization of Activities in a Recurrent Network
• Spike-timing-dependent Hebbian plasticity as temporal difference learning
• Predictive learning in rate-coded neuronal networks: a theoretical approach towards classical conditioning
• Fading memory and time series prediction in recurrent networks with different forms of plasticity

Cites work

• A simplified neuron model as a principal component analyzer
• Averaging methods in nonlinear dynamical systems
• Bayesian Spiking Neurons I: Inference
• Fading memory and time series prediction in recurrent networks with different forms of plasticity
• Inverse Problems in Neural Field Theory
• Multiscale analysis of slow-fast neuronal learning models with noise
• Nonlinear systems
• Recognizing recurrent neural networks (rRNN): Bayesian inference for recurrent neural net\-works
• Relating STDP to BCM
• Reservoir computing approaches to recurrent neural network training
• Spike-timing-dependent Hebbian plasticity as temporal difference learning
• Spiking Neuron Models
• Why spikes? Hebbian learning and retrieval of time-resolved excitation patterns

Cited in

(8)

• Distributed synaptic weights in a LIF neural network and learning rules
• Slow feature analysis with spiking neurons and its application to audio stimuli
• Evaluating the extent to which homeostatic plasticity learns to compute prediction errors in unstructured neuronal networks
• Fading memory and time series prediction in recurrent networks with different forms of plasticity
• A local echo state property through the largest Lyapunov exponent
• Dynamic branching in a neural network model for probabilistic prediction of sequences
• Nonlinear partial differential equations in neuroscience: from modeling to mathematical theory
• Relative entropy minimizing noisy non-linear neural network to approximate stochastic processes

This page was built for publication: A biological gradient descent for prediction through a combination of STDP and homeostatic plasticity

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5378278)