论文标题
低活动性监督卷积尖峰神经网络应用于语音命令识别
Low-activity supervised convolutional spiking neural networks applied to speech commands recognition
论文作者
论文摘要
深神经网络（DNN）是许多与语音相关的任务中的当前最新模型。但是，人们对更现实，硬件友好和节能的模型（称为Spiking Neural Networks（SNN））的兴趣越来越大。最近，已经表明，可以使用替代梯度技巧的反向传播以有监督的方式对SNN进行有效的培训。在这项工作中，我们使用有监督的SNN报告语音命令（SC）识别实验。我们探索了此任务的泄漏 - 综合发生（LIF）神经元模型，并表明由堆叠的卷积卷积尖峰层组成的模型可以达到与Google SC V1数据集上的标准DNN相近的错误率：5.5％：同时保持非常稀疏的尖峰活动，低于5％的5％，感谢新的常规化定期。我们还表明，建模神经元膜电位的泄漏非常有用，因为LIF模型的表现明显优于其非裸模型。
Deep Neural Networks (DNNs) are the current state-of-the-art models in many speech related tasks. There is a growing interest, though, for more biologically realistic, hardware friendly and energy efficient models, named Spiking Neural Networks (SNNs). Recently, it has been shown that SNNs can be trained efficiently, in a supervised manner, using backpropagation with a surrogate gradient trick. In this work, we report speech command (SC) recognition experiments using supervised SNNs. We explored the Leaky-Integrate-Fire (LIF) neuron model for this task, and show that a model comprised of stacked dilated convolution spiking layers can reach an error rate very close to standard DNNs on the Google SC v1 dataset: 5.5%, while keeping a very sparse spiking activity, below 5%, thank to a new regularization term. We also show that modeling the leakage of the neuron membrane potential is useful, since the LIF model outperformed its non-leaky model counterpart significantly.
