Transformer has shown tremendous progress in Automatic Speech Recognition (ASR), outperforming recurrent neural network-based approaches. Transformer architecture is good at parallelizing data to accelerate as well as capturing content-based global interaction. However, most studies with Transfomer have been utilized only shallow features extracted from the backbone without taking advantage of the deep feature that possesses invariant property. In this paper, we propose a novel framework with two streams that consist of different resolution spectrograms for each steam aiming to capture both shallow and deep features. The feature extraction module consists of a deep network for small resolution spectrogram and a shallow network for large resolution spectrogram. The backbone obtains not only detailed acoustic information for speech-text alignment but also sentence invariant features such as speaker information. Both features are fused with our proposed fusion method and then input into the Transformer encoder-decoder. With our method, the proposed framework shows competitive performance on Mandarin corpus. It outperforms various current state-of-the-art results on the HKUST Mandarian telephone ASR benchmark with a CER of 21.08. To the best of our knowledge, this is the first investigation of incorporating deep features to the backbone.
Recently, directly utilize raw waveforms as input is widely explored for the speaker verification system. For example, RawNet [1] and RawNet2 [2] extract feature embeddings from raw waveforms, which largely reduce the front-end computation and achieve state-of-the-art performance. However, they do not consider the speech speed influence which is different from person to person. In this paper, we propose a novel finite-difference network to obtain speaker embeddings. It incorporates speaker speech speed by computing the finite difference between adjacent time speech pieces. Furthermore, we design a hierarchical layer to capture multiscale speech speed features to improve the system accuracy. The speaker embeddings is then input into the GRU to aggregate utterance-level features before the softmax loss. Experiment results on official VoxCeleb1 test data and expanded evaluation on VoxCeleb1-E and VoxCeleb-H protocols show our method outperforms existing state-of-the-art systems. To facilitate further research, code is available at https://github.com/happyjin/FDN
Unsupervised deep learning has recently demonstrated the promise to produce high-quality samples. While it has tremendous potential to promote the image colorization task, the performance is limited owing to the manifold hypothesis in machine learning. This study presents a novel scheme that exploiting the score-based generative model in wavelet domain to address the issue. By taking advantage of the multi-scale and multi-channel representation via wavelet transform, the proposed model learns the priors from stacked wavelet coefficient components, thus learns the image characteristics under coarse and detail frequency spectrums jointly and effectively. Moreover, such a highly flexible generative model without adversarial optimization can execute colorization tasks better under dual consistency terms in wavelet domain, namely data-consistency and structure-consistency. Specifically, in the training phase, a set of multi-channel tensors consisting of wavelet coefficients are used as the input to train the network by denoising score matching. In the test phase, samples are iteratively generated via annealed Langevin dynamics with data and structure consistencies. Experiments demonstrated remarkable improvements of the proposed model on colorization quality, particularly on colorization robustness and diversity.
Although significant progress in automatic learning of steganographic cost has been achieved recently, existing methods designed for spatial images are not well applicable to JPEG images which are more common media in daily life. The difficulties of migration mostly lie in the unique and complicated JPEG characteristics caused by 8x8 DCT mode structure. To address the issue, in this paper we extend an existing automatic cost learning scheme to JPEG, where the proposed scheme called JEC-RL (JPEG Embedding Cost with Reinforcement Learning) is explicitly designed to tailor the JPEG DCT structure. It works with the embedding action sampling mechanism under reinforcement learning, where a policy network learns the optimal embedding policies via maximizing the rewards provided by an environment network. The policy network is constructed following a domain-transition design paradigm, where three modules including pixel-level texture complexity evaluation, DCT feature extraction, and mode-wise rearrangement, are proposed. These modules operate in serial, gradually extracting useful features from a decompressed JPEG image and converting them into embedding policies for DCT elements, while considering JPEG characteristics including inter-block and intra-block correlations simultaneously. The environment network is designed in a gradient-oriented way to provide stable reward values by using a wide architecture equipped with a fixed preprocessing layer with 8x8 DCT basis filters. Extensive experiments and ablation studies demonstrate that the proposed method can achieve good security performance for JPEG images against both advanced feature based and modern CNN based steganalyzers.
Recently, the witness of the rapidly growing popularity of short videos on different Internet platforms has intensified the need for a background music (BGM) retrieval system. However, existing video-music retrieval methods only based on the visual modality cannot show promising performance regarding videos with fine-grained virtual contents. In this paper, we also investigate the widely added voice-overs in short videos and propose a novel framework to retrieve BGM for fine-grained short videos. In our framework, we use the self-attention (SA) and the cross-modal attention (CMA) modules to explore the intra- and the inter-relationships of different modalities respectively. For balancing the modalities, we dynamically assign different weights to the modal features via a fusion gate. For paring the query and the BGM embeddings, we introduce a triplet pseudo-label loss to constrain the semantics of the modal embeddings. As there are no existing virtual-content video-BGM retrieval datasets, we build and release two virtual-content video datasets HoK400 and CFM400. Experimental results show that our method achieves superior performance and outperforms other state-of-the-art methods with large margins.
In the standard data analysis framework, data is first collected (once for all), and then data analysis is carried out. With the advancement of digital technology, decisionmakers constantly analyze past data and generate new data through the decisions they make. In this paper, we model this as a Markov decision process and show that the dynamic interaction between data generation and data analysis leads to a new type of bias -- reinforcement bias -- that exacerbates the endogeneity problem in standard data analysis. We propose a class of instrument variable (IV)-based reinforcement learning (RL) algorithms to correct for the bias and establish their asymptotic properties by incorporating them into a two-timescale stochastic approximation framework. A key contribution of the paper is the development of new techniques that allow for the analysis of the algorithms in general settings where noises feature time-dependency. We use the techniques to derive sharper results on finite-time trajectory stability bounds: with a polynomial rate, the entire future trajectory of the iterates from the algorithm fall within a ball that is centered at the true parameter and is shrinking at a (different) polynomial rate. We also use the technique to provide formulas for inferences that are rarely done for RL algorithms. These formulas highlight how the strength of the IV and the degree of the noise's time dependency affect the inference.
Matching module plays a critical role in display advertising systems. Without query from user, it is challenging for system to match user traffic and ads suitably. System packs up a group of users with common properties such as the same gender or similar shopping interests into a crowd. Here term crowd can be viewed as a tag over users. Then advertisers bid for different crowds and deliver their ads to those targeted users. Matching module in most industrial display advertising systems follows a two-stage paradigm. When receiving a user request, matching system (i) finds the crowds that the user belongs to; (ii) retrieves all ads that have targeted those crowds. However, in applications such as display advertising at Alibaba, with very large volumes of crowds and ads, both stages of matching have to truncate the long-tailed parts for online serving, under limited latency. That's to say, not all ads have the chance to participate in online matching. This results in sub-optimal result for both advertising performance and platform revenue. In this paper, we study the truncation problem and propose a Truncation Free Matching System (TFMS). The basic idea is to decouple the matching computation from the online pipeline. Instead of executing the two-stage matching when user visits, TFMS utilizes a near-line truncation-free matching to pre-calculate and store those top valuable ads for each user. Then the online pipeline just needs to fetch the pre-stored ads as matching results. In this way, we can jump out of online system's latency and computation cost limitations, and leverage flexible computation resource to finish the user-ad matching. TFMS has been deployed in our productive system since 2019, bringing (i) more than 50% improvement of impressions for advertisers who encountered truncation before, (ii) 9.4% Revenue Per Mile gain, which is significant enough for the business.
This paper proposes an iterative generative model for solving the automatic colorization problem. Although previous researches have shown the capability to generate plausible color, the edge color overflow and the requirement of the reference images still exist. The starting point of the unsupervised learning in this study is the observation that the gradient map possesses latent information of the image. Therefore, the inference process of the generative modeling is conducted in joint intensity-gradient domain. Specifically, a set of intensity-gradient formed high-dimensional tensors, as the network input, are used to train a powerful noise conditional score network at the training phase. Furthermore, the joint intensity-gradient constraint in data-fidelity term is proposed to limit the degree of freedom within generative model at the iterative colorization stage, and it is conducive to edge-preserving. Extensive experiments demonstrated that the system outperformed state-of-the-art methods whether in quantitative comparisons or user study.
For e-commerce platforms such as Taobao and Amazon, advertisers play an important role in the entire digital ecosystem: their behaviors explicitly influence users' browsing and shopping experience; more importantly, advertiser's expenditure on advertising constitutes a primary source of platform revenue. Therefore, providing better services for advertisers is essential for the long-term prosperity for e-commerce platforms. To achieve this goal, the ad platform needs to have an in-depth understanding of advertisers in terms of both their marketing intents and satisfaction over the advertising performance, based on which further optimization could be carried out to service the advertisers in the correct direction. In this paper, we propose a novel Deep Satisfaction Prediction Network (DSPN), which models advertiser intent and satisfaction simultaneously. It employs a two-stage network structure where advertiser intent vector and satisfaction are jointly learned by considering the features of advertiser's action information and advertising performance indicators. Experiments on an Alibaba advertisement dataset and online evaluations show that our proposed DSPN outperforms state-of-the-art baselines and has stable performance in terms of AUC in the online environment. Further analyses show that DSPN not only predicts advertisers' satisfaction accurately but also learns an explainable advertiser intent, revealing the opportunities to optimize the advertising performance further.
Bipartite b-matching is fundamental in algorithm design, and has been widely applied into economic markets, labor markets, etc. These practical problems usually exhibit two distinct features: large-scale and dynamic, which requires the matching algorithm to be repeatedly executed at regular intervals. However, existing exact and approximate algorithms usually fail in such settings due to either requiring intolerable running time or too much computation resource. To address this issue, we propose \texttt{NeuSearcher} which leverages the knowledge learned from previously instances to solve new problem instances. Specifically, we design a multichannel graph neural network to predict the threshold of the matched edges weights, by which the search region could be significantly reduced. We further propose a parallel heuristic search algorithm to iteratively improve the solution quality until convergence. Experiments on both open and industrial datasets demonstrate that \texttt{NeuSearcher} can speed up 2 to 3 times while achieving exactly the same matching solution compared with the state-of-the-art approximation approaches.