Maximum a posteriori adaptation of network parameters in deep models

By Zhen Huang, Sabato Marco Siniscalchi, I-Fan Chen, Jinyu Li, Jiadong Wu, Chin-Hui Lee. 2015

54 citations recorded by the source corpus.

We present a Bayesian approach to adapting parameters of a well-trained context-dependent, deep-neural-network, hidden Markov model (CD-DNN-HMM) to improve automatic speech recognition performance. Given an abundance of DNN parameters but with only a limited amount of data, the effectiveness of the adapted DNN model can often be compromised. We formulate maximum a posteriori (MAP) adaptation of parameters of a specially designed CD-DNN-HMM with an augmented linear hidden networks connected to the output tied states, or senones, and compare it to feature space MAP linear regression previously proposed. Experimental evidences on the 20,000-word open vocabulary Wall Street Journal task demonstrate the feasibility of the proposed framework. In supervised adaptation, the proposed MAP adaptation approach provides more than 10% relative error reduction and consistently outperforms the conventional transformation based methods. Furthermore, we present an initial attempt to generate hierarchical priors to improve adaptation efficiency and effectiveness with limited adaptation data by exploiting similarities among senones.

Subfields: Artificial Intelligence, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Jinyu Li
2. Yifan Gong
3. Dong Yu
4. Zhong Meng
5. Chin-Hui Lee
6. Li Deng
7. Sabato Marco Siniscalchi
8. Zhen Huang

In Signal Processing

1. Jinyu Li
2. Yifan Gong
3. Zhong Meng
4. Dong Yu
5. Chin-Hui Lee
6. Zhen Huang
7. Sabato Marco Siniscalchi
8. Zhuo Chen

Citations in the corpus

• Overcoming Catastrophic Forgetting by Incremental Moment Matching (2017, methodology): “applied sequential Bayesian to adapt a deep neural network to the specific user in the speech recognition domain [15, 16].”
• Speaker-Invariant Training Via Adversarial Learning (2018, background): “Therefore, speaker adaptation meth-ods are widely used to boost the recognition system performance [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13].”
• Recent Advances in End-to-End Automatic Speech Recognition (2021, methodology): “The first one is using regularization techniques such as Kullback-Leibler (KL) divergence regularization [169], maximum a posteriori adaptation [170], or elastic weight consolidation [171], to ensure the adapted models do not overfit the limited adaptation data [160, 162–164].”
• Hierarchical Bayesian combination of plug-in maximum a posteriori decoders in deep neural networks-based speech recognition and speaker adaptation (2017, methodology): “For adaptation of the DNN based ASR system in this paper, an affine transform network, namely the LHN, is inserted into the original neural architecture (Gemello et al., 2007) and MAP LHN adaption follows (Huang et al., 2015b).”
• Unsupervised adaptation with domain separation networks for robust speech recognition (2017, methodology): “To address this issue, regularization-based approaches are proposed in [9, 10, 11, 12] to regularize the neuron output distributions or the model parameters.”