Compression of Acoustic Event Detection Models with Low-rank Matrix Factorization and Quantization Training

Bowen Shi; Ming Sun; Chieh-Chi Kao; Viktor Rozgic; Spyros Matsoukas; Chao Wang

Compression of Acoustic Event Detection Models with Low-rank Matrix Factorization and Quantization Training

Audio and Speech Processing 2019-05-03 v1 Computation and Language Sound

Authors: Bowen Shi , Ming Sun , Chieh-Chi Kao , Viktor Rozgic , Spyros Matsoukas , Chao Wang

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.

Keywords

deep learning for detection and classification audio signal processing matrix optimization and algorithms

Cite

@article{arxiv.1905.00855,
  title  = {Compression of Acoustic Event Detection Models with Low-rank Matrix Factorization and Quantization Training},
  author = {Bowen Shi and Ming Sun and Chieh-Chi Kao and Viktor Rozgic and Spyros Matsoukas and Chao Wang},
  journal= {arXiv preprint arXiv:1905.00855},
  year   = {2019}
}

Comments

NeuralPS 2018 CDNNRIA workshop

Compression of Acoustic Event Detection Models with Low-rank Matrix Factorization and Quantization Training

Abstract

Keywords

Cite

Comments

Related papers