The training of deep neural networks (DNNs) is currently predominantly done using first-order methods. Some of these methods (e.g., Adam, AdaGrad, and RMSprop, and their variants) incorporate a small amount of curvature information by using a diagonal matrix to precondition the stochastic gradient. Recently, effective second-order methods, such as KFAC, K-BFGS, Shampoo, and TNT, have been developed for training DNNs, by preconditioning the stochastic gradient by layer-wise block-diagonal matrices. Here we propose and analyze the convergence of an approximate natural gradient method, mini-block Fisher (MBF), that lies in between these two classes of methods. Specifically, our method uses a block-diagonal approximation to the Fisher matrix, where for each layer in the DNN, whether it is convolutional or feed-forward and fully connected, the associated diagonal block is also block-diagonal and is composed of a large number of mini-blocks of modest size. Our novel approach utilizes the parallelism of GPUs to efficiently perform computations on the large number of matrices in each layer. Consequently, MBF's per-iteration computational cost is only slightly higher than it is for first-order methods. Finally, the performance of our proposed method is compared to that of several baseline methods, on both Auto-encoder and CNN problems, to validate its effectiveness both in terms of time efficiency and generalization power.
Unlike the conventional facial expressions, micro-expressions are involuntary and transient facial expressions capable of revealing the genuine emotions that people attempt to hide. Therefore, they can provide important information in a broad range of applications such as lie detection, criminal detection, etc. Since micro-expressions are transient and of low intensity, however, their detection and recognition is difficult and relies heavily on expert experiences. Due to its intrinsic particularity and complexity, video-based micro-expression analysis is attractive but challenging, and has recently become an active area of research. Although there have been numerous developments in this area, thus far there has been no comprehensive survey that provides researchers with a systematic overview of these developments with a unified evaluation. Accordingly, in this survey paper, we first highlight the key differences between macro- and micro-expressions, then use these differences to guide our research survey of video-based micro-expression analysis in a cascaded structure, encompassing the neuropsychological basis, datasets, features, spotting algorithms, recognition algorithms, applications and evaluation of state-of-the-art approaches. For each aspect, the basic techniques, advanced developments and major challenges are addressed and discussed. Furthermore, after considering the limitations of existing micro-expression datasets, we present and release a new dataset - called micro-and-macro expression warehouse (MMEW) - containing more video samples and more labeled emotion types. We then perform a unified comparison of representative methods on CAS(ME)2 for spotting, and on MMEW and SAMM for recognition, respectively. Finally, some potential future research directions are explored and outlined.
Multi-speaker singing voice synthesis is to generate the singing voice sung by different speakers. To generalize to new speakers, previous zero-shot singing adaptation methods obtain the timbre of the target speaker with a fixed-size embedding from single reference audio. However, they face several challenges: 1) the fixed-size speaker embedding is not powerful enough to capture full details of the target timbre; 2) single reference audio does not contain sufficient timbre information of the target speaker; 3) the pitch inconsistency between different speakers also leads to a degradation in the generated voice. In this paper, we propose a new model called MR-SVS to tackle these problems. Specifically, we employ both a multi-reference encoder and a fixed-size encoder to encode the timbre of the target speaker from multiple reference audios. The Multi-reference encoder can capture more details and variations of the target timbre. Besides, we propose a well-designed pitch shift method to address the pitch inconsistency problem. Experiments indicate that our method outperforms the baseline method both in naturalness and similarity.
High-fidelity multi-singer singing voice synthesis is challenging for neural vocoder due to the singing voice data shortage, limited singer generalization, and large computational cost. Existing open corpora could not meet requirements for high-fidelity singing voice synthesis because of the scale and quality weaknesses. Previous vocoders have difficulty in multi-singer modeling, and a distinct degradation emerges when conducting unseen singer singing voice generation. To accelerate singing voice researches in the community, we release a large-scale, multi-singer Chinese singing voice dataset OpenSinger. To tackle the difficulty in unseen singer modeling, we propose Multi-Singer, a fast multi-singer vocoder with generative adversarial networks. Specifically, 1) Multi-Singer uses a multi-band generator to speed up both training and inference procedure. 2) to capture and rebuild singer identity from the acoustic feature (i.e., mel-spectrogram), Multi-Singer adopts a singer conditional discriminator and conditional adversarial training objective. 3) to supervise the reconstruction of singer identity in the spectrum envelopes in frequency domain, we propose an auxiliary singer perceptual loss. The joint training approach effectively works in GANs for multi-singer voices modeling. Experimental results verify the effectiveness of OpenSinger and show that Multi-Singer improves unseen singer singing voices modeling in both speed and quality over previous methods. The further experiment proves that combined with FastSpeech 2 as the acoustic model, Multi-Singer achieves strong robustness in the multi-singer singing voice synthesis pipeline. Samples are available at https://Multi-Singer.github.io/
Nowadays, recommender systems already impact almost every facet of peoples lives. To provide personalized high quality recommendation results, conventional systems usually train pointwise rankers to predict the absolute value of objectives and leverage a distinct shallow tower to estimate and alleviate the impact of position bias. However, with such a training paradigm, the optimization target differs a lot from the ranking metrics valuing the relative order of top ranked items rather than the prediction precision of each item. Moreover, as the existing system tends to recommend more relevant items at higher positions, it is difficult for the shallow tower based methods to precisely attribute the user feedback to the impact of position or relevance. Therefore, there exists an exciting opportunity for us to get enhanced performance if we manage to solve the aforementioned issues. Unbiased learning to rank algorithms, which are verified to model the relative relevance accurately based on noisy feedback, are appealing candidates and have already been applied in many applications with single categorical labels, such as user click signals. Nevertheless, the existing unbiased LTR methods cannot properly handle multiple feedback incorporating both categorical and continuous labels. Accordingly, we design a novel unbiased LTR algorithm to tackle the challenges, which innovatively models position bias in the pairwise fashion and introduces the pairwise trust bias to separate the position bias, trust bias, and user relevance explicitly. Experiment results on public benchmark datasets and internal live traffic show the superior results of the proposed method for both categorical and continuous labels.
High-fidelity singing voice synthesis is challenging for neural vocoders due to extremely long continuous pronunciation, high sampling rate and strong expressiveness. Existing neural vocoders designed for text-to-speech cannot directly be applied to singing voice synthesis because they result in glitches in the generated spectrogram and poor high-frequency reconstruction. To tackle the difficulty of singing modeling, in this paper, we propose SingGAN, a singing voice vocoder with generative adversarial network. Specifically, 1) SingGAN uses source excitation to alleviate the glitch problem in the spectrogram; and 2) SingGAN adopts multi-band discriminators and introduces frequency-domain loss and sub-band feature matching loss to supervise high-frequency reconstruction. To our knowledge, SingGAN is the first vocoder designed towards high-fidelity multi-speaker singing voice synthesis. Experimental results show that SingGAN synthesizes singing voices with much higher quality (0.41 MOS gains) over the previous method. Further experiments show that combined with FastSpeech~2 as an acoustic model, SingGAN achieves high robustness in the singing voice synthesis pipeline and also performs well in speech synthesis.
Semantically connecting users and items is a fundamental problem for the matching stage of an industrial recommender system. Recent advances in this topic are based on multi-channel retrieval to efficiently measure users' interest on items from the massive candidate pool. However, existing work are primarily built upon pre-defined retrieval channels, including User-CF (U2U), Item-CF (I2I), and Embedding-based Retrieval (U2I), thus access to the limited correlation between users and items which solely entail from partial information of latent interactions. In this paper, we propose a model-agnostic integrated cross-channel (MIC) approach for the large-scale recommendation, which maximally leverages the inherent multi-channel mutual information to enhance the matching performance. Specifically, MIC robustly models correlation within user-item, user-user, and item-item from latent interactions in a universal schema. For each channel, MIC naturally aligns pairs with semantic similarity and distinguishes them otherwise with more uniform anisotropic representation space. While state-of-the-art methods require specific architectural design, MIC intuitively considers them as a whole by enabling the complete information flow among users and items. Thus MIC can be easily plugged into other retrieval recommender systems. Extensive experiments show that our MIC helps several state-of-the-art models boost their performance on two real-world benchmarks. The satisfactory deployment of the proposed MIC on industrial online services empirically proves its scalability and flexibility.
Federated learning enables collaborative training of machine learning models under strict privacy restrictions and federated text-to-speech aims to synthesize natural speech of multiple users with a few audio training samples stored in their devices locally. However, federated text-to-speech faces several challenges: very few training samples from each speaker are available, training samples are all stored in local device of each user, and global model is vulnerable to various attacks. In this paper, we propose a novel federated learning architecture based on continual learning approaches to overcome the difficulties above. Specifically, 1) we use gradual pruning masks to isolate parameters for preserving speakers' tones; 2) we apply selective masks for effectively reusing knowledge from tasks; 3) a private speaker embedding is introduced to keep users' privacy. Experiments on a reduced VCTK dataset demonstrate the effectiveness of FedSpeech: it nearly matches multi-task training in terms of multi-speaker speech quality; moreover, it sufficiently retains the speakers' tones and even outperforms the multi-task training in the speaker similarity experiment.