Abstract:The architecture design and multi-scale learning principles of the human brain that evolved over hundreds of millions of years are crucial to realizing human-like intelligence. Spiking Neural Network (SNN) based Liquid State Machine (LSM) serves as a suitable architecture to study brain-inspired intelligence because of its brain-inspired structure and the potential for integrating multiple biological principles. Existing researches on LSM focus on different certain perspectives, including high-dimensional encoding or optimization of the liquid layer, network architecture search, and application to hardware devices. There is still a lack of in-depth inspiration from the learning and structural evolution mechanism of the brain. Considering these limitations, this paper presents a novel LSM learning model that integrates adaptive structural evolution and multi-scale biological learning rules. For structural evolution, an adaptive evolvable LSM model is developed to optimize the neural architecture design of liquid layer with separation property. For brain-inspired learning of LSM, we propose a dopamine-modulated Bienenstock-Cooper-Munros (DA-BCM) method that incorporates global long-term dopamine regulation and local trace-based BCM synaptic plasticity. Comparative experimental results on different decision-making tasks show that introducing structural evolution of the liquid layer, and the DA-BCM regulation of the liquid layer and the readout layer could improve the decision-making ability of LSM and flexibly adapt to rule reversal. This work is committed to exploring how evolution can help to design more appropriate network architectures and how multi-scale neuroplasticity principles coordinated to enable the optimization and learning of LSMs for relatively complex decision-making tasks.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the limitations in the structure and function of the existing Liquid State Machine (LSM) models. Specifically: 1. **Structural limitations**: In most existing LSM models, the structure of the liquid layer is usually fixed after initialization and is only used as a high - dimensional encoding method, lacking the ability to dynamically adapt to task changes. This fixed structure limits the LSM's ability to handle complex tasks. 2. **Functional limitations**: Existing LSM research mainly focuses on high - dimensional encoding or optimization of the liquid layer, network architecture search, and application in hardware devices, but lacks in - depth reference to the brain's learning and structural evolution mechanisms. For example, many models use the gradient - descent method to train the read - out layer without training the liquid layer, which is different from the real learning mechanism of the brain. In addition, some methods that attempt to train the liquid layer through local synaptic plasticity (such as STDP or Hebb rules) are also limited to simple tasks. To overcome these limitations, this paper proposes a new LSM learning model that combines adaptive structure evolution and multi - scale biological learning rules. Specific contributions include: - **Adaptive structure evolution**: An adaptively evolving LSM model has been developed to optimize the neural architecture design of the liquid layer, making it have better separation characteristics. - **Biologically - inspired learning**: A dopamine - modulated Bienenstock - Cooper - Munro (DA - BCM) method has been proposed, which combines global long - term dopamine modulation and local trace - based BCM synaptic plasticity. Through these improvements, the paper aims to improve the performance of LSM in complex decision - making tasks and enable it to flexibly adapt to rule changes. Experimental results show that the introduction of structural evolution of the liquid layer and DA - BCM modulation can significantly improve the decision - making ability and adaptability of LSM.

Adaptive structure evolution and biologically plausible synaptic plasticity for recurrent spiking neural networks

Adaptive structure evolution and biologically plausible synaptic plasticity for recurrent spiking neural networks

Emergence of brain-inspired small-world spiking neural network through neuroevolution

Surrogate-Assisted Evolutionary Search of Spiking Neural Architectures in Liquid State Machines

Increasing Liquid State Machine Performance with Edge-of-Chaos Dynamics Organized by Astrocyte-modulated Plasticity

Surrogate-Assisted Cooperative Co-evolutionary Reservoir Architecture Search for Liquid State Machines

Biologically Inspired Structure Learning with Reverse Knowledge Distillation for Spiking Neural Networks

Evolutionary Optimization Of Liquid State Machines For Robust Learning

Adaptive Multi-Level Firing for Direct Training Deep Spiking Neural Networks

A Brain-Inspired Spiking Neural Network Model with Temporal Encoding and Learning

Toward Efficient Processing and Learning with Spikes: New Approaches for Multispike Learning

Enhancing learning in spiking neural networks through neuronal heterogeneity and neuromodulatory signaling

Learning the Plasticity: Plasticity-Driven Learning Framework in Spiking Neural Networks

Spike-driven Multi-Scale Learning with Hybrid Mechanisms of Spiking Dendrites.

Multi-scale Evolutionary Neural Architecture Search for Deep Spiking Neural Networks

Adaptive Synaptic Scaling in Spiking Networks for Continual Learning and Enhanced Robustness

Two-compartment neuronal spiking model expressing brain-state specific apical-amplification, -isolation and -drive regimes

Incorporating structural plasticity into self-organization recurrent networks for sequence learning

Synaptic transistor with multiple biological functions based on metal-organic frameworks combined with the LIF model of a spiking neural network to recognize temporal information

A Progressive Training Framework for Spiking Neural Networks with Learnable Multi-hierarchical Model

SPikE-SSM: A Sparse, Precise, and Efficient Spiking State Space Model for Long Sequences Learning