Jianwei Zhang

Peng Fan, Changhao Shan, Sining Sun et al.

Recently, Conformer as a backbone network for end-to-end automatic speech recognition achieved state-of-the-art performance. The Conformer block leverages a self-attention mechanism to capture global information, along with a convolutional neural network to capture local information, resulting in improved performance. However, the Conformer-based model encounters an issue with the self-attention mechanism, as computational complexity grows quadratically with the length of the input sequence. Inspired by previous Connectionist Temporal Classification (CTC) guided blank skipping during decoding, we introduce intermediate CTC outputs as guidance into the downsampling procedure of the Conformer encoder. We define the frame with non-blank output as key frame. Specifically, we introduce the key frame-based self-attention (KFSA) mechanism, a novel method to reduce the computation of the self-attention mechanism using key frames. The structure of our proposed approach comprises two encoders. Following the initial encoder, we introduce an intermediate CTC loss function to compute the label frame, enabling us to extract the key frames and blank frames for KFSA. Furthermore, we introduce the key frame-based downsampling (KFDS) mechanism to operate on high-dimensional acoustic features directly and drop the frames corresponding to blank labels, which results in new acoustic feature sequences as input to the second encoder. By using the proposed method, which achieves comparable or higher performance than vanilla Conformer and other similar work such as Efficient Conformer. Meantime, our proposed method can discard more than 60\% useless frames during model training and inference, which will accelerate the inference speed significantly. This work code is available in {https://github.com/scufan1990/Key-Frame-Mechanism-For-Efficient-Conformer}

Tianying Ji, Yu Luo, Fuchun Sun et al.

Learning high-quality $Q$-value functions plays a key role in the success of many modern off-policy deep reinforcement learning (RL) algorithms. Previous works primarily focus on addressing the value overestimation issue, an outcome of adopting function approximators and off-policy learning. Deviating from the common viewpoint, we observe that $Q$-values are often underestimated in the latter stage of the RL training process, potentially hindering policy learning and reducing sample efficiency. We find that such a long-neglected phenomenon is often related to the use of inferior actions from the current policy in Bellman updates as compared to the more optimal action samples in the replay buffer. To address this issue, our insight is to incorporate sufficient exploitation of past successes while maintaining exploration optimism. We propose the Blended Exploitation and Exploration (BEE) operator, a simple yet effective approach that updates $Q$-value using both historical best-performing actions and the current policy. Based on BEE, the resulting practical algorithm BAC outperforms state-of-the-art methods in over 50 continuous control tasks and achieves strong performance in failure-prone scenarios and real-world robot tasks. Benchmark results and videos are available at https://jity16.github.io/BEE/.

Weijie Kong, Qi Tian, Zijian Zhang et al. · tencent-ai, tsinghua

Recent advancements in video generation have significantly impacted daily life for both individuals and industries. However, the leading video generation models remain closed-source, resulting in a notable performance gap between industry capabilities and those available to the public. In this report, we introduce HunyuanVideo, an innovative open-source video foundation model that demonstrates performance in video generation comparable to, or even surpassing, that of leading closed-source models. HunyuanVideo encompasses a comprehensive framework that integrates several key elements, including data curation, advanced architectural design, progressive model scaling and training, and an efficient infrastructure tailored for large-scale model training and inference. As a result, we successfully trained a video generative model with over 13 billion parameters, making it the largest among all open-source models. We conducted extensive experiments and implemented a series of targeted designs to ensure high visual quality, motion dynamics, text-video alignment, and advanced filming techniques. According to evaluations by professionals, HunyuanVideo outperforms previous state-of-the-art models, including Runway Gen-3, Luma 1.6, and three top-performing Chinese video generative models. By releasing the code for the foundation model and its applications, we aim to bridge the gap between closed-source and open-source communities. This initiative will empower individuals within the community to experiment with their ideas, fostering a more dynamic and vibrant video generation ecosystem. The code is publicly available at https://github.com/Tencent/HunyuanVideo.

Kaixin Bai, Huajian Zeng, Lei Zhang et al.

Transparent object depth perception poses a challenge in everyday life and logistics, primarily due to the inability of standard 3D sensors to accurately capture depth on transparent or reflective surfaces. This limitation significantly affects depth map and point cloud-reliant applications, especially in robotic manipulation. We developed a vision transformer-based algorithm for stereo depth recovery of transparent objects. This approach is complemented by an innovative feature post-fusion module, which enhances the accuracy of depth recovery by structural features in images. To address the high costs associated with dataset collection for stereo camera-based perception of transparent objects, our method incorporates a parameter-aligned, domain-adaptive, and physically realistic Sim2Real simulation for efficient data generation, accelerated by AI algorithm. Our experimental results demonstrate the model's exceptional Sim2Real generalizability in real-world scenarios, enabling precise depth mapping of transparent objects to assist in robotic manipulation. Project details are available at https://sites.google.com/view/cleardepth/ .

Ruowei Wang, Jiaqi Li, Dan Zeng et al.

Generating high-quality meshes with complex structures and realistic surfaces is the primary goal of 3D generative models. Existing methods typically employ sequence data or deformable tetrahedral grids for mesh generation. However, sequence-based methods have difficulty producing complex structures with many faces due to memory limits. The deformable tetrahedral grid-based method MeshDiffusion fails to recover realistic surfaces due to the inherent ambiguity in deformable grids. We propose the GenUDC framework to address these challenges by leveraging the Unsigned Dual Contouring (UDC) as the mesh representation. UDC discretizes a mesh in a regular grid and divides it into the face and vertex parts, recovering both complex structures and fine details. As a result, the one-to-one mapping between UDC and mesh resolves the ambiguity problem. In addition, GenUDC adopts a two-stage, coarse-to-fine generative process for 3D mesh generation. It first generates the face part as a rough shape and then the vertex part to craft a detailed shape. Extensive evaluations demonstrate the superiority of UDC as a mesh representation and the favorable performance of GenUDC in mesh generation. The code and trained models are available at https://github.com/TrepangCat/GenUDC.

Xiaoyun Zhang, Xieyi Ping, Jianwei Zhang

Previous work optimizes traditional active learning (AL) processes with incremental neural network architecture search (Active-iNAS) based on data complexity change, which improves the accuracy and learning efficiency. However, Active-iNAS trains several models and selects the model with the best generalization performance for querying the subsequent samples after each active learning cycle. The independent training processes lead to an insufferable computational budget, which is significantly inefficient and limits search flexibility and final performance. To address this issue, we propose a novel active strategy with the method called structured variational inference (SVI) or structured neural depth search (SNDS) whereby we could use the gradient descent method in neural network depth search during AL processes. At the same time, we theoretically demonstrate that the current VI-based methods based on the mean-field assumption could lead to poor performance. We apply our strategy using three querying techniques and three datasets and show that our strategy outperforms current methods.

Xincheng Yu, Dongyue Guo, Jianwei Zhang et al.

Radio speech echo is a specific phenomenon in the air traffic control (ATC) domain, which degrades speech quality and further impacts automatic speech recognition (ASR) accuracy. In this work, a time-domain recognition-oriented speech enhancement (ROSE) framework is proposed to improve speech intelligibility and also advance ASR accuracy based on convolutional encoder-decoder-based U-Net framework, which serves as a plug-and-play tool in ATC scenarios and does not require additional retraining of the ASR model. Specifically, 1) In the U-Net architecture, an attention-based skip-fusion (ABSF) module is applied to mine shared features from encoders using an attention mask, which enables the model to effectively fuse the hierarchical features. 2) A channel and sequence attention (CSAtt) module is innovatively designed to guide the model to focus on informative features in dual parallel attention paths, aiming to enhance the effective representations and suppress the interference noises. 3) Based on the handcrafted features, ASR-oriented optimization targets are designed to improve recognition performance in the ATC environment by learning robust feature representations. By incorporating both the SE-oriented and ASR-oriented losses, ROSE is implemented in a multi-objective learning manner by optimizing shared representations across the two task objectives. The experimental results show that the ROSE significantly outperforms other state-of-the-art methods for both the SE and ASR tasks, in which all the proposed improvements are confirmed by designed experiments. In addition, the proposed approach can contribute to the desired performance improvements on public datasets.

Pengju Zhang, Chenxi Sun, Jianwei Zhang et al.

We have developed an individual identification system based on magnetocardiography (MCG) signals captured using optically pumped magnetometers (OPMs). Our system utilizes pattern recognition to analyze the signals obtained at different positions on the body, by scanning the matrices composed of MCG signals with a 2*2 window. In order to make use of the spatial information of MCG signals, we transform the signals from adjacent small areas into four channels of a dataset. We further transform the data into time-frequency matrices using wavelet transforms and employ a convolutional neural network (CNN) for classification. As a result, our system achieves an accuracy rate of 97.04% in identifying individuals. This finding indicates that the MCG signal holds potential for use in individual identification systems, offering a valuable tool for personalized healthcare management.

Dongyue Guo, Zheng Zhang, Zhen Yan et al.

Flight Trajectory Prediction (FTP) is an essential task in Air Traffic Control (ATC), which can assist air traffic controllers in managing airspace more safely and efficiently. Existing approaches generally perform multi-horizon FTP tasks in an autoregressive manner, thereby suffering from error accumulation and low-efficiency problems. In this paper, a novel framework, called FlightBERT++, is proposed to i) forecast multi-horizon flight trajectories directly in a non-autoregressive way, and ii) improve the limitation of the binary encoding (BE) representation in the FlightBERT framework. Specifically, the proposed framework is implemented by a generalized encoder-decoder architecture, in which the encoder learns the temporal-spatial patterns from historical observations and the decoder predicts the flight status for the future horizons. Compared to conventional architecture, an innovative horizon-aware contexts generator is dedicatedly designed to consider the prior horizon information, which further enables non-autoregressive multi-horizon prediction. Additionally, the Gray code representation and the differential prediction paradigm are designed to cope with the high-bit misclassifications of the BE representation, which significantly reduces the outliers in the predictions. Moreover, a differential prompted decoder is proposed to enhance the capability of the differential predictions by leveraging the stationarity of the differential sequence. Extensive experiments are conducted to validate the proposed framework on a real-world flight trajectory dataset. The experimental results demonstrated that the proposed framework outperformed the competitive baselines in both FTP performance and computational efficiency.

Peng Fan, Dongyue Guo, Yi Lin et al.

In this work, we propose a new automatic speech recognition (ASR) system based on feature learning and an end-to-end training procedure for air traffic control (ATC) systems. The proposed model integrates the feature learning block, recurrent neural network (RNN), and connectionist temporal classification loss to build an end-to-end ASR model. Facing the complex environments of ATC speech, instead of the handcrafted features, a learning block is designed to extract informative features from raw waveforms for acoustic modeling. Both the SincNet and 1D convolution blocks are applied to process the raw waveforms, whose outputs are concatenated to the RNN layers for the temporal modeling. Thanks to the ability to learn representations from raw waveforms, the proposed model can be optimized in a complete end-to-end manner, i.e., from waveform to text. Finally, the multilingual issue in the ATC domain is also considered to achieve the ASR task by constructing a combined vocabulary of Chinese characters and English letters. The proposed approach is validated on a multilingual real-world corpus (ATCSpeech), and the experimental results demonstrate that the proposed approach outperforms other baselines, achieving a 6.9\% character error rate.

Dongyue Guo, Jianwei Zhang, Bo Yang et al.

Automatic spoken instruction understanding (SIU) of the controller-pilot conversations in the air traffic control (ATC) requires not only recognizing the words and semantics of the speech but also determining the role of the speaker. However, few of the published works on the automatic understanding systems in air traffic communication focus on speaker role identification (SRI). In this paper, we formulate the SRI task of controller-pilot communication as a binary classification problem. Furthermore, the text-based, speech-based, and speech and text based multi-modal methods are proposed to achieve a comprehensive comparison of the SRI task. To ablate the impacts of the comparative approaches, various advanced neural network architectures are applied to optimize the implementation of text-based and speech-based methods. Most importantly, a multi-modal speaker role identification network (MMSRINet) is designed to achieve the SRI task by considering both the speech and textual modality features. To aggregate modality features, the modal fusion module is proposed to fuse and squeeze acoustic and textual representations by modal attention mechanism and self-attention pooling layer, respectively. Finally, the comparative approaches are validated on the ATCSpeech corpus collected from a real-world ATC environment. The experimental results demonstrate that all the comparative approaches are worked for the SRI task, and the proposed MMSRINet shows the competitive performance and robustness than the other methods on both seen and unseen data, achieving 98.56%, and 98.08% accuracy, respectively.

Yi Lin, Bo Yang, Linchao Li et al.

In this paper, a multilingual end-to-end framework, called as ATCSpeechNet, is proposed to tackle the issue of translating communication speech into human-readable text in air traffic control (ATC) systems. In the proposed framework, we focus on integrating the multilingual automatic speech recognition (ASR) into one model, in which an end-to-end paradigm is developed to convert speech waveform into text directly, without any feature engineering or lexicon. In order to make up for the deficiency of the handcrafted feature engineering caused by ATC challenges, a speech representation learning (SRL) network is proposed to capture robust and discriminative speech representations from the raw wave. The self-supervised training strategy is adopted to optimize the SRL network from unlabeled data, and further to predict the speech features, i.e., wave-to-feature. An end-to-end architecture is improved to complete the ASR task, in which a grapheme-based modeling unit is applied to address the multilingual ASR issue. Facing the problem of small transcribed samples in the ATC domain, an unsupervised approach with mask prediction is applied to pre-train the backbone network of the ASR model on unlabeled data by a feature-to-feature process. Finally, by integrating the SRL with ASR, an end-to-end multilingual ASR framework is formulated in a supervised manner, which is able to translate the raw wave into text in one model, i.e., wave-to-text. Experimental results on the ATCSpeech corpus demonstrate that the proposed approach achieves a high performance with a very small labeled corpus and less resource consumption, only 4.20% label error rate on the 58-hour transcribed corpus. Compared to the baseline model, the proposed approach obtains over 100% relative performance improvement which can be further enhanced with the increasing of the size of the transcribed samples.

Jianwei Zhang, zhenxing Wang, yuhui Zheng et al.

With the development of the super-resolution convolutional neural network (SRCNN), deep learning technique has been widely applied in the field of image super-resolution. Previous works mainly focus on optimizing the structure of SRCNN, which have been achieved well performance in speed and restoration quality for image super-resolution. However, most of these approaches only consider a specific scale image during the training process, while ignoring the relationship between different scales of images. Motivated by this concern, in this paper, we propose a cascaded convolution neural network for image super-resolution (CSRCNN), which includes three cascaded Fast SRCNNs and each Fast SRCNN can process a specific scale image. Images of different scales can be trained simultaneously and the learned network can make full use of the information resided in different scales of images. Extensive experiments show that our network can achieve well performance for image SR.

Lin Cong, Michael Görner, Philipp Ruppel et al.

Planar pushing remains a challenging research topic, where building the dynamic model of the interaction is the core issue. Even an accurate analytical dynamic model is inherently unstable because physics parameters such as inertia and friction can only be approximated. Data-driven models usually rely on large amounts of training data, but data collection is time consuming when working with real robots. In this paper, we collect all training data in a physics simulator and build an LSTM-based model to fit the pushing dynamics. Domain Randomization is applied to capture the pushing trajectories of a generalized class of objects. When executed on the real robot, the trained recursive model adapts to the tracked object's real dynamics within a few steps. We propose the algorithm \emph{Recurrent} Model Predictive Path Integral (RMPPI) as a variation of the original MPPI approach, employing state-dependent recurrent models. As a comparison, we also train a Deep Deterministic Policy Gradient (DDPG) network as a model-free baseline, which is also used as the action generator in the data collection phase. During policy training, Hindsight Experience Replay is used to improve exploration efficiency. Pushing experiments on our UR5 platform demonstrate the model's adaptability and the effectiveness of the proposed framework.

Xu Zou, Qiuye Jia, Jianwei Zhang et al.

Graph Convolution Networks (GCNs) are becoming more and more popular for learning node representations on graphs. Though there exist various developments on sampling and aggregation to accelerate the training process and improve the performances, limited works focus on dealing with the dimensional information imbalance of node representations. To bridge the gap, we propose a method named Dimensional reweighting Graph Convolution Network (DrGCN). We theoretically prove that our DrGCN can guarantee to improve the stability of GCNs via mean field theory. Our dimensional reweighting method is very flexible and can be easily combined with most sampling and aggregation techniques for GCNs. Experimental results demonstrate its superior performances on several challenging transductive and inductive node classification benchmark datasets. Our DrGCN also outperforms existing models on an industrial-sized Alibaba recommendation dataset.

Yukuo Cen, Xu Zou, Jianwei Zhang et al.

Network embedding (or graph embedding) has been widely used in many real-world applications. However, existing methods mainly focus on networks with single-typed nodes/edges and cannot scale well to handle large networks. Many real-world networks consist of billions of nodes and edges of multiple types, and each node is associated with different attributes. In this paper, we formalize the problem of embedding learning for the Attributed Multiplex Heterogeneous Network and propose a unified framework to address this problem. The framework supports both transductive and inductive learning. We also give the theoretical analysis of the proposed framework, showing its connection with previous works and proving its better expressiveness. We conduct systematical evaluations for the proposed framework on four different genres of challenging datasets: Amazon, YouTube, Twitter, and Alibaba. Experimental results demonstrate that with the learned embeddings from the proposed framework, we can achieve statistically significant improvements (e.g., 5.99-28.23% lift by F1 scores; p<<0.01, t-test) over previous state-of-the-art methods for link prediction. The framework has also been successfully deployed on the recommendation system of a worldwide leading e-commerce company, Alibaba Group. Results of the offline A/B tests on product recommendation further confirm the effectiveness and efficiency of the framework in practice.

Yunpei Zheng, Lin Li, Luo Zhong et al.

User profiling means exploiting the technology of machine learning to predict attributes of users, such as demographic attributes, hobby attributes, preference attributes, etc. It's a powerful data support of precision marketing. Existing methods mainly study network behavior, personal preferences, post texts to build user profile. Through our data analysis of micro-blog, we find that females show more positive and have richer emotions than males in online social platform. This difference is very conducive to the distinction between genders. Therefore, we argue that sentiment context is important as well for user profiling.This paper focuses on exploiting microblog user posts to predict one of the demographic labels: gender. We propose a Sentiment Representation Learning based Multi-Layer Perceptron(SRL-MLP) model to classify gender. First we build a sentiment polarity classifier in advance by training Long Short-Term Memory(LSTM) model on e-commerce review corpus. Next we transfer sentiment representation to a basic MLP network. Last we conduct experiments on gender classification by sentiment representation. Experimental results show that our approach can improve gender classification accuracy by 5.53\%, from 84.20\% to 89.73\%.