Gaussian Process Neural Networks for Speech Recognition
文献类型：会议论文
| 作者 | Max W. Y. Lam; Shoukang Hu; Xurong Xie; Shansong Liu; Jianwei Yu; Rongfeng Su; Xunying Liu; Helen Meng |
| 出版日期 | 2018 |
| 会议日期 | 2018 |
| 会议地点 | HYDERABAD, INDIA |
| 英文摘要 | Deep neural networks (DNNs) play an important role in state of-the-art speech recognition systems. One important issue as sociated with DNNs and artificial neural networks in general is the selection of suitable model structures, for example, the form of hidden node activation functions to use. Due to lack of automatic model selection techniques, the choice of acti vation functions has been largely empirically based. In addi tion, the use of deterministic, fixed-point parameter estimates is prone to over-fitting when given limited training data. In order to model both models structural and parametric uncertainty, a novel form of DNN architecture using non-parametric activa tion functions based on Gaussian process (GP), Gaussian pro cess neural networks (GPNN), is proposed in this paper. Initial experiments conducted on the ARPA Resource Management task suggest that the proposed GPNN acoustic models outper formed the baseline sigmoid activation based DNN by 3.40% to 24.25% relatively in terms of word error rate. Consistent per formance improvements over the DNN baseline were also ob tained by varying the number of hidden nodes and the number of spectral basis functions |
| 源URL | [http://ir.siat.ac.cn:8080/handle/172644/13716]  |
| 专题 | 深圳先进技术研究院_集成所 |
| 推荐引用方式 GB/T 7714 | Max W. Y. Lam,Shoukang Hu,Xurong Xie,et al. Gaussian Process Neural Networks for Speech Recognition[C]. 见:. HYDERABAD, INDIA. 2018. |
入库方式： OAI收割
来源：深圳先进技术研究院
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。