Neuroscience inspired scientific machine learning (Part-1): Variable spiking neuron for regression
This addresses energy efficiency and regression capability in spiking neural networks, but appears incremental as it blends existing neuron types.
The paper tackles the problem of redundant information transfer in neural networks, which increases complexity and power consumption, by introducing a Variable Spiking Neuron (VSN) that reduces redundant firing and is suitable for regression tasks while keeping energy low.
Redundant information transfer in a neural network can increase the complexity of the deep learning model, thus increasing its power consumption. We introduce in this paper a novel spiking neuron, termed Variable Spiking Neuron (VSN), which can reduce the redundant firing using lessons from biological neuron inspired Leaky Integrate and Fire Spiking Neurons (LIF-SN). The proposed VSN blends LIF-SN and artificial neurons. It garners the advantage of intermittent firing from the LIF-SN and utilizes the advantage of continuous activation from the artificial neuron. This property of the proposed VSN makes it suitable for regression tasks, which is a weak point for the vanilla spiking neurons, all while keeping the energy budget low. The proposed VSN is tested against both classification and regression tasks. The results produced advocate favorably towards the efficacy of the proposed spiking neuron, particularly for regression tasks.