Energy-efficient artificial pyramidal neuron dendrites for visual information inference

In the human brain, a specialized type of pyramidal neuron dendrite exhibits nonlinear spike integration and sparse parallel processing capabilities. This unique structure enables efficient execution of visual tasks by activating only a small subset of neurons, playing a crucial role in high-level information inference. However, traditional neuromorphic devices are typically modeled as node-based signal integrators, requiring the activation of all neurons for perception. This approach struggles to effectively replicate the highly efficient spatiotemporal information processing of biological dendrites. In this study, we present an artificial pyramidal neuron dendrite (APND) array that integrates pyramidal neurons, synapses, and dendrites, emulating the spatiotemporal spike integration properties of biological pyramidal neurons for precise parallel computation. Through multi-gate threshold regulation, dendrites enable parallel sparse spiking inference with random spatial distribution. This inference process forms a sparse dendritic spiking neural network (SD-SNN) that can perform compression, recognition, and prediction. As a result, the SD-SNN achieves high-efficiency static and dynamic object processing while using only 0.5% of neurons. This reduces the number of ADCs by 99.5% and decreases power consumption by 98% and 65%, respectively. Our work reduces neural activity in the perception process by 99.5% while enhancing spatiotemporal computing capabilities and computational efficiency.