- Bayesian learning; Bayesian signal processing (MLR-BAYL)
- Bounds on performance (MLR-PERF)
- Applications in Systems Biology (MLR-SYSB)
- Applications in Music and Audio Processing (MLR-MUSI)
- Applications in Data Fusion (MLR-FUSI)
- Cognitive information processing (MLR-COGP)
- Distributed and Cooperative Learning (MLR-DIST)
- Learning theory and algorithms (MLR-LEAR)
- Neural network learning (MLR-NNLR)
- Information-theoretic learning (MLR-INFO)
- Independent component analysis (MLR-ICAN)
- Graphical and kernel methods (MLR-GRKN)
- Other applications of machine learning (MLR-APPL)
- Pattern recognition and classification (MLR-PATT)
- Source separation (MLR-SSEP)
- Sequential learning; sequential decision methods (MLR-SLER)
We propose a novel adversarial speaker adaptation (ASA) scheme, in which adversarial learning is applied to regularize the distribution of deep hidden features in a speaker-dependent (SD) deep neural network (DNN) acoustic model to be close to that of a fixed speaker-independent (SI) DNN acoustic model during adaptation. An additional discriminator network is introduced to distinguish the deep features generated by the SD model from those produced by the SI model.
Adversarial domain-invariant training (ADIT) proves to be effective in suppressing the effects of domain variability in acoustic modeling and has led to improved performance in automatic speech recognition (ASR). In ADIT, an auxiliary domain classifier takes in equally-weighted deep features from a deep neural network (DNN) acoustic model and is trained to improve their domain-invariance by optimizing an adversarial loss function.
The use of deep networks to extract embeddings for speaker recognition has proven successfully. However, such embeddings are susceptible to performance degradation due to the mismatches among the training, enrollment, and test conditions. In this work, we propose an adversarial speaker verification (ASV) scheme to learn the condition-invariant deep embedding via adversarial multi-task training. In ASV, a speaker classification network and a condition identification network are jointly optimized to minimize the speaker classification loss and simultaneously mini-maximize the condition loss.
The teacher-student (T/S) learning has been shown to be effective for a variety of problems such as domain adaptation and model compression. One shortcoming of the T/S learning is that a teacher model, not always perfect, sporadically produces wrong guidance in form of posterior probabilities that misleads the student model towards a suboptimal performance.
Tensor decomposition is a powerful tool for analyzing multiway data. Nowadays, with the fast development of multisensor technology, more and more data appear in higher-order (order >= 4) and nonnegative form. However, the decomposition of higher-order nonnegative tensor suffers from poor convergence and low speed. In this study, we propose a new nonnegative CANDECOM/PARAFAC (NCP) model using proximal algorithm. The block principal pivoting method in alternating nonnegative least squares (ANLS) framework is employed to minimize the objective function.