Abstract
In this paper, we present a compression approach based on the combination of low-rank matrix factorization and quantization training, to reduce complexity for neural network based acoustic event detection (AED) models. Our experimental results show this combined compression approach is very effective. For a three-layer long short-term memory (LSTM) based AED model, the original model size can be reduced to 1% with negligible loss of accuracy. Our approach enables the feasibility of deploying AED for resource-constraint applications.
Abstract (translated)
本文提出了一种基于低阶矩阵分解和量化训练相结合的压缩方法,以降低基于神经网络的声事件检测(AED)模型的复杂度。实验结果表明,这种组合压缩方法是非常有效的。对于基于三层长短期存储器(LSTM)的AED模型,可以将原始模型的大小减少到1%,并且精度损失可以忽略不计。我们的方法使部署AED用于资源约束应用的可行性成为可能。
URL
https://arxiv.org/abs/1905.00855