新闻动态

团队在2024年6月6日《自然·电子》上发表了类脑神经计算模型 “Dendristor”。

  这种创新的树突网络模拟了树突状结构及其固有的时空处理特性，为未来人工智能提供了高能效的视觉感知能力。Dendristor 的突出之处在于它处理信息的方式与神经元及其网络的生物形态非常相似，而不是目前人工神经网络典型的批处理方式。Dendristor模型实现了树突分支间和神经元间的特定塑性，从而提高稀疏神经网络中的学习效率。这种方法允许Dendristor在其树突分支内对传入信号的序列和方向进行编码，从而提高其识别运动的能力。特别是模型中包含的“沉默突触”，即由树突分支电位激活的突触，增强了其对信号方向的敏感性，优化了其视觉感知过程。这种创新的树突计算利用了独特的方法，在人工智能、神经计算和脑启发式计算领域开辟了新的可能性。

Neuromorphic dendritic network computation with silent synapses for visual motion perception

Eunhye Baek, Sen Song, Chang-Ki Baek, Zhao Rong, Luping Shi & Carlo Vittorio Cannistraci

Abstract：

       Neuromorphic technologies typically employ a point neuron model, neglecting the spatiotemporal nature of neuronal computation. Dendritic morphology and synaptic organization are structurally tailored for spatiotemporal information processing, such as visual perception. Here we report a neuromorphic computational model that integrates synaptic organization with dendritic tree-like morphology. Based on the physics of multigate silicon nanowire transistors with ion-doped sol–gel films, our model—termed dendristor—performs dendritic computation at the device and neural-circuit level. The dendristor offers the bioplausible nonlinear integration of excitatory/inhibitory synaptic inputs and silent synapses with diverse spatial distribution dependency, emulating direction selectivity, which is the feature that reacts to signal direction on the dendrite. We also develop a neuromorphic dendritic neural circuit—a network of interconnected dendritic neurons—that serves as a building block for the design of a multilayer network system that emulates three-dimensional spatial motion perception in the retina.