Adaptive learning by extremal dynamics and negative feedback

Abstract

We describe a mechanism for biological learning and adaptation based on two simple principles: (I) Neuronal activity propagates only through the network's strongest synaptic connections (extremal dynamics), and (II) The strengths of active synapses are reduced if mistakes are made, otherwise no changes occur (negative feedback). The balancing of those two tendencies typically shapes a synaptic landscape with configurations which are barely stable, and therefore highly flexible. This allows for swift adaptation to new situations. Recollection of past successes is achieved by punishing synapses which have once participated in activity associated with successful outputs much less than neurons that have never been successful. Despite its simplicity, the model can readily learn to solve complicated nonlinear tasks, even in the presence of noise. In particular, the learning time for the benchmark parity problem scales algebraically with the problem size N, with an exponent $k\sim 1.4$.

References in corpus (2)

• Extremal Optimization: Methods derived from Co-Evolution
• Beyond Hebb: Exclusive-OR and Biological Learning

Cited by in corpus (26)

• Emergent complex neural dynamics
• Colloquium: Criticality and dynamical scaling in living systems
• Critical brain networks
• Learning as a phenomenon occurring in a critical state
• Energy-efficient stochastic computing with superparamagnetic tunnel junctions
• Self-organization toward criticality by synaptic plasticity
• The brain: What is critical about it?
• Avalanches in self-organized critical neural networks: A minimal model for the neural SOC universality class
• Single-Neuron Criticality Optimizes Analog Dendritic Computation
• Optimal percentage of inhibitory synapses in multi-task learning
• Critical neural networks with short and long term plasticity
• Deterministic excitable media under Poisson drive: power law responses, spiral waves and dynamic range
• Self-organized criticality in neural networks from activity-based rewiring
• Neuronal avalanches of a self-organized neural network with active-neuron-dominant structure
• Intelligent systems in the context of surrounding environment
• Memory and burstiness in dynamic networks
• Learning by mistakes in memristor networks
• Spatial features of synaptic adaptation affecting learning performance
• Order-disorder transition in the Chialvo-Bak `minibrain' controlled by network geometry
• The collective brain is critical
• Adaptivity and `Per learning'
• Mechanisms of self-organized quasicriticality in neuronal networks models
• Emergent complexity: what uphill analysis or downhill invention can not do
• Biologically inspired learning in a layered neural net
• Self-organized annealing in laterally inhibited neural networks shows power law decay
• SWAF: Swarm Algorithm Framework for Numerical Optimization