Adaptive Plasticity in Spiking Neuron Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing plasticity implementations in spiking neural networks do not effectively adapt to changing input characteristics, limiting network behavior and learning performance.
Innovation Solution
The implementation of adaptive plasticity mechanisms that dynamically adjust connection efficacies based on similarity measures between neuron outputs and inputs, using methods such as cross-correlograms and STDP rules to determine long-term potentiation and depression components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing plasticity implementations are used in spiking neural networks, then the network can operate with fixed connection efficacies, but the network cannot effectively adapt to changing input characteristics
Solution Approach 1:
The patent implements dynamic plasticity mechanisms where connection efficacies are not fixed but continuously adjusted based on spike timing relationships. The STDP (spike-timing-dependent plasticity) rule dynamically modifies synaptic weights according to the temporal order of pre-synaptic and post-synaptic spikes, enabling the network to adapt to changing input characteristics while maintaining a relatively simple implementation framework
Solution Approach 2:
The patent changes the parameter of connection efficacy dynamically through plasticity mechanisms. By modifying synaptic weights based on spike timing correlations and similarity measures, the system enables adaptability to varying input patterns without requiring complete reconfiguration of the network architecture, thus balancing adaptability improvement with controlled complexity increase
2Productivity
If adaptive plasticity mechanisms are implemented to improve learning performance, then convergence speed increases, but computational complexity increases
Solution Approach 1:
The patent applies partial action by implementing plasticity updates only when specific conditions are met (e.g., when spike timing correlations exceed thresholds or when similarity measures indicate meaningful patterns). This selective updating approach accelerates convergence by focusing computational resources on critical learning moments while avoiding unnecessary computations during stable or redundant states, thus improving productivity without proportionally increasing overall computational complexity
3Reliability
If fixed connection efficacies are used, then the network structure remains simple, but the network cannot learn temporally stable patterns
Solution Approach 1:
The patent implements feedback mechanisms where the network monitors its own spike timing patterns and uses this information to adjust connection efficacies. The STDP rule provides continuous feedback based on whether pre-synaptic spikes precede or follow post-synaptic spikes, enabling the network to learn temporally stable patterns by reinforcing causally relevant connections while pruning spurious correlations, thus improving reliability through structured feedback rather than complex ad-hoc adjustments
Data Source
AI summary
Apparatus and methods for plasticity in a spiking neuron network. In one implementation, a plasticity mechanism is configured based on a similarity measure between neuron post-synaptic and pre-synaptic activity. The similarity measure may comprise a cross-correlogram between the output spike train and input spike train, determined over a plasticity window. Several correlograms, corresponding to individual input connections delivering pre-synaptic input, may be combined. The combination may comprise for example a weighted average. The averaged correlograms may be used to construct the long term potentiation component of the plasticity. The long term depression component of the plasticity may comprise e.g., a monotonic function based on a statistical parameter associated with the adaptively determined long term potentiation component.