Developmental plasticity plays a prominent role in shaping the brain's structure during ongoing learning in response to dynamically changing environments. However, the existing network compression methods for deep artificial neural networks (ANNs) and spiking neural networks (SNNs) draw little inspiration from the brain's developmental plasticity mechanisms, thus limiting their ability to learn efficiently, rapidly, and accurately. This paper proposed a developmental plasticity-inspired adaptive pruning (DPAP) method, with inspiration from the adaptive developmental pruning of dendritic spines, synapses, and neurons according to the "use it or lose it, gradually decay" principle. The proposed DPAP model considers multiple biologically realistic mechanisms (such as dendritic spine dynamic plasticity, activity-dependent neural spiking trace, and local synaptic plasticity), with the addition of an adaptive pruning strategy, so that the network structure can be dynamically optimized during learning without any pre-training and retraining. We demonstrated that the proposed DPAP method applied to deep ANNs and SNNs could learn efficient network architectures. Extensive comparative experiments show consistent and remarkable performance and speed boost with the extremely compressed networks on a diverse set of benchmark tasks, especially neuromorphic datasets for SNNs. This work explores how developmental plasticity enables the complex deep networks to gradually evolve into brain-like efficient and compact structures, eventually achieving state-of-the-art (SOTA) performance for biologically realistic SNNs.
翻译:发展型可塑性在根据动态变化的环境不断学习的过程中,在形成大脑结构的过程中发挥着突出作用;然而,现有的深层人工神经网络网络网络压缩方法(ANNS)和神经网络(SNNS)从大脑的发育型可塑性机制中几乎没有什么灵感,从而限制了其高效、快速和准确地学习的能力。本文件建议了一种由发育型可塑性激发的适应性调整调整(DPAP)方法,其灵感来自根据“使用或失去它,逐渐衰败”原则对登地脊椎、突触和神经系统的适应性发育剪裁剪裁(DPAP)方法。拟议的DPP模型认为多种生物性现实性机制(例如斜度脊椎动态可塑性、活动性神经性闪烁性跟踪和局部合成性合成性造影性),加上适应性裁剪裁战略,以便在不经过任何培训和再培训的情况下,在学习过程中能够动态地优化网络结构。我们证明,根据“使用或失去它,逐渐衰败”原则,拟议的DPP方法可以学习高效的网络结构结构。拟议的DPP模式认为多种生物现实性现实性机制(例如硬性软质的S-S-SIMLS-S-S-S-rode-rode-rode-rode-rode-traal St-traal-traal-traal-traal-traal-traal-traal-tra-lax-lax-lax-traal-lax-traction-tra-trax-lax-lax-lax-traal-traal-traal-traveld-lax-s-laxSil-trax-travel-trax-tra-trax-trax-trax-trax-trax-trax-trax-tra-tra-trax-trax-trax-trax-trax-trax-tra-tra-tra-trax-trax-trax-trax-trax-trax-s-s-s-sl-s-s-s-lax-s-s-s-s-s-s-s-trax-s-s-s-s-S-S-S-S-S-S-S-S-S-S-S-S-S-