arXiv subjects

Yong Xu

Publications and source records attributed to Yong Xu.

At least 145 records · Page 8Linked to original sources

arXiv2024

LibriheavyMix: A 20,000-Hour Dataset for Single-Channel Reverberant Multi-Talker Speech Separation, ASR and Speaker Diarization

The evolving speech processing landscape is increasingly focused on complex scenarios like meetings or cocktail parties with multiple simultaneous speakers and far-field conditions. Existing methodologies for addressing these challenges fall into two categories: multi-channel and single-channel solutions. Single-channel approaches, notable for their generality and convenience, do not require specific information about microphone arrays. This paper presents a large-scale far-field overlapping speech dataset, crafted to advance research in speech separation, recognition, and speaker diarization. This dataset is a critical resource for decoding ``Who said What and When'' in multi-talker, reverberant environments, a daunting challenge in the field. Additionally, we introduce a pipeline system encompassing speech separation, recognition, and diarization as a foundational benchmark. Evaluations on the WHAMR! dataset validate the broad applicability of the proposed data.

cs.SD↗

arXiv2024

Advancing Multi-talker ASR Performance with Large Language Models

Recognizing overlapping speech from multiple speakers in conversational scenarios is one of the most challenging problem for automatic speech recognition (ASR). Serialized output training (SOT) is a classic method to address multi-talker ASR, with the idea of concatenating transcriptions from multiple speakers according to the emission times of their speech for training. However, SOT-style transcriptions, derived from concatenating multiple related utterances in a conversation, depend significantly on modeling long contexts. Therefore, compared to traditional methods that primarily emphasize encoder performance in attention-based encoder-decoder (AED) architectures, a novel approach utilizing large language models (LLMs) that leverages the capabilities of pre-trained decoders may be better suited for such complex and challenging scenarios. In this paper, we propose an LLM-based SOT approach for multi-talker ASR, leveraging pre-trained speech encoder and LLM, fine-tuning them on multi-talker dataset using appropriate strategies. Experimental results demonstrate that our approach surpasses traditional AED-based methods on the simulated dataset LibriMix and achieves state-of-the-art performance on the evaluation set of the real-world dataset AMI, outperforming the AED model trained with 1000 times more supervised data in previous works.

eess.AS↗

arXiv2024

Dynamical Transition due to Feedback-induced Skin Effect

The traditional dynamical phase transition refers to the appearance of singularities in an observable with respect to a control parameter for a late-time state or singularities in the rate function of the Loschmidt echo with respect to time. Here, we study the many-body dynamics in a continuously monitored free fermion system with conditional feedback under open boundary conditions. We surprisingly find a novel dynamical transition from a logarithmic scaling of the entanglement entropy to an area-law scaling as time evolves. The transition, which is noticeably different from the conventional dynamical phase transition, arises from the competition between the bulk dynamics and boundary skin effects. In addition, we find that while quasidisorder or disorder cannot drive a transition for the steady state, a transition occurs for the maximum entanglement entropy during the time evolution, which agrees well with the entanglement transition for the steady state of the dynamics under periodic boundary conditions.

quant-ph↗

arXiv2024

OV-DQUO: Open-Vocabulary DETR with Denoising Text Query Training and Open-World Unknown Objects Supervision

Open-vocabulary detection aims to detect objects from novel categories beyond the base categories on which the detector is trained. However, existing open-vocabulary detectors trained on base category data tend to assign higher confidence to trained categories and confuse novel categories with the background. To resolve this, we propose OV-DQUO, an \textbf{O}pen-\textbf{V}ocabulary DETR with \textbf{D}enoising text \textbf{Q}uery training and open-world \textbf{U}nknown \textbf{O}bjects supervision. Specifically, we introduce a wildcard matching method. This method enables the detector to learn from pairs of unknown objects recognized by the open-world detector and text embeddings with general semantics, mitigating the confidence bias between base and novel categories. Additionally, we propose a denoising text query training strategy. It synthesizes foreground and background query-box pairs from open-world unknown objects to train the detector through contrastive learning, enhancing its ability to distinguish novel objects from the background. We conducted extensive experiments on the challenging OV-COCO and OV-LVIS benchmarks, achieving new state-of-the-art results of 45.6 AP50 and 39.3 mAP on novel categories respectively, without the need for additional training data. Models and code are released at \url{https://github.com/xiaomoguhz/OV-DQUO}

cs.CV↗

arXiv2024

OpenCity: Open Spatio-Temporal Foundation Models for Traffic Prediction

Accurate traffic forecasting is crucial for effective urban planning and transportation management, enabling efficient resource allocation and enhanced travel experiences. However, existing models often face limitations in generalization, struggling with zero-shot prediction on unseen regions and cities, as well as diminished long-term accuracy. This is primarily due to the inherent challenges in handling the spatial and temporal heterogeneity of traffic data, coupled with the significant distribution shift across time and space. In this work, we aim to unlock new possibilities for building versatile, resilient and adaptive spatio-temporal foundation models for traffic prediction. To achieve this goal, we introduce a novel foundation model, named OpenCity, that can effectively capture and normalize the underlying spatio-temporal patterns from diverse data characteristics, facilitating zero-shot generalization across diverse urban environments. OpenCity integrates the Transformer architecture with graph neural networks to model the complex spatio-temporal dependencies in traffic data. By pre-training OpenCity on large-scale, heterogeneous traffic datasets, we enable the model to learn rich, generalizable representations that can be seamlessly applied to a wide range of traffic forecasting scenarios. Experimental results demonstrate that OpenCity exhibits exceptional zero-shot predictive performance. Moreover, OpenCity showcases promising scaling laws, suggesting the potential for developing a truly one-for-all traffic prediction solution that can adapt to new urban contexts with minimal overhead. We made our proposed OpenCity model open-source and it is available at the following link: https://github.com/HKUDS/OpenCity.

cs.LG↗

arXiv2024

Transfer learning relaxation, electronic structure and continuum model for twisted bilayer MoTe$_2$

Large-scale moiré systems are extraordinarily sensitive, with even minute atomic shifts leading to significant changes in electronic structures. Here, we investigate the lattice relaxation effect on moiré band structures in twisted bilayer MoTe$_2$ with two approaches: (a) large-scale plane-wave basis first principle calculation down to $2.88^{\circ}$, (b) transfer learning structure relaxation + local-basis first principles calculation down to $1.1^{\circ}$. We use two types of van der Waals corrections: the D2 method of Grimme and the density-dependent energy correction, and find that the density-dependent energy correction yields a continuous evolution of bandwidth with twist angles. Based on the above results. we develop a more complete continuum model with a single set of parameters for a wide range of twist angles, and perform many-body simulations at $ν=-1,-2/3, -1/3$.

cond-mat.str-el↗

arXiv2024

Neural-network Density Functional Theory Based on Variational Energy Minimization

Deep-learning density functional theory (DFT) shows great promise to significantly accelerate material discovery and potentially revolutionize materials research. However, current research in this field primarily relies on data-driven supervised learning, making the developments of neural networks and DFT isolated from each other. In this work, we present a theoretical framework of neural-network DFT, which unifies the optimization of neural networks with the variational computation of DFT, enabling physics-informed unsupervised learning. Moreover, we develop a differential DFT code incorporated with deep-learning DFT Hamiltonian, and introduce algorithms of automatic differentiation and backpropagation into DFT, demonstrating the capability of neural-network DFT. The physics-informed neural-network architecture not only surpasses conventional approaches in accuracy and efficiency, but also offers a new paradigm for developing deep-learning DFT methods.

physics.comp-ph↗

arXiv2024

Fractional Chern insulators in moiré flat bands with high Chern numbers

Recent discoveries of zero-field fractional Chern insulators in moiré materials have attracted intensive research interests. However, most current theoretical and experimental attempts focus on systems with low Chern number bands, in analogy to the Landau levels. Here we propose candidate material systems for realizing fractional Chern insulators with higher Chern numbers. The material setup involves $Γ$-valley twisted homobilayer transition metal dichalcogenides in proximity to a skyrmion lattice. The skyrmion exchange potential induces a flat band with a high Chern number $C = -2$. Using the momentum-space projected exact diagonalization method, we perform a comprehensive study at various filling factors, confirming the generalized Jain series. Our research provides theoretical guidance on realizing unconventional fractional Chern insulators beyond the Landau level picture.

cond-mat.str-el↗

arXiv2024

Resonant Reheating

We investigate a novel reheating scenario proceeding through $s$-channel inflaton annihilation, mediated by a massive scalar. If the inflaton $ϕ$ oscillates around the minimum of a monomial potential $\propto ϕ^{n}$, we reveal the emergence of resonance phenomena originating from the dynamic evolution of the inflaton mass for $n>2$. Consequently, a resonance appears in both the radiation and the temperature evolution during the reheating process. By solving the coupled Boltzmann equations, we present solutions for radiation and temperature. We find non-trivial temperature characteristics during reheating, depending on the value of $n$ and the masses of the inflaton and mediator. Some phenomenological aspects of the model are explored. As a concrete example, we show that the same mediator participates in the genesis of dark matter, modifying the standard freeze-in dynamics. In addition, we demonstrate that the resonant reheating scenario could be tested by next-generation low- and high-frequency gravitational wave detectors.

hep-ph↗

arXiv2024

Tracking Meets LoRA: Faster Training, Larger Model, Stronger Performance

Motivated by the Parameter-Efficient Fine-Tuning (PEFT) in large language models, we propose LoRAT, a method that unveils the power of large ViT model for tracking within laboratory-level resources. The essence of our work lies in adapting LoRA, a technique that fine-tunes a small subset of model parameters without adding inference latency, to the domain of visual tracking. However, unique challenges and potential domain gaps make this transfer not as easy as the first intuition. Firstly, a transformer-based tracker constructs unshared position embedding for template and search image. This poses a challenge for the transfer of LoRA, usually requiring consistency in the design when applied to the pre-trained backbone, to downstream tasks. Secondly, the inductive bias inherent in convolutional heads diminishes the effectiveness of parameter-efficient fine-tuning in tracking models. To overcome these limitations, we first decouple the position embeddings in transformer-based trackers into shared spatial ones and independent type ones. The shared embeddings, which describe the absolute coordinates of multi-resolution images (namely, the template and search images), are inherited from the pre-trained backbones. In contrast, the independent embeddings indicate the sources of each token and are learned from scratch. Furthermore, we design an anchor-free head solely based on MLP to adapt PETR, enabling better performance with less computational overhead. With our design, 1) it becomes practical to train trackers with the ViT-g backbone on GPUs with only memory of 25.8GB (batch size of 16); 2) we reduce the training time of the L-224 variant from 35.0 to 10.8 GPU hours; 3) we improve the LaSOT SUC score from 0.703 to 0.742 with the L-224 variant; 4) we fast the inference speed of the L-224 variant from 52 to 119 FPS. Code and models are available at https://github.com/LitingLin/LoRAT.

cs.CV↗

arXiv2024

Deep learning density functional theory Hamiltonian in real space

Deep learning electronic structures from ab initio calculations holds great potential to revolutionize computational materials studies. While existing methods proved success in deep-learning density functional theory (DFT) Hamiltonian matrices, they are limited to DFT programs using localized atomic-like bases and heavily depend on the form of the bases. Here, we propose the DeepH-r method for deep-learning DFT Hamiltonians in real space, facilitating the prediction of DFT Hamiltonian in a basis-independent manner. An equivariant neural network architecture for modeling the real-space DFT potential is developed, targeting a more fundamental quantity in DFT. The real-space potential exhibits simplified principles of equivariance and enhanced nearsightedness, further boosting the performance of deep learning. When applied to evaluate the Hamiltonian matrix, this method significantly improved in accuracy, as exemplified in multiple case studies. Given the abundance of data in the real-space potential, this work may pave a novel pathway for establishing a ``large materials model" with increased accuracy.

physics.comp-ph↗

arXiv2024

3D Shape Completion on Unseen Categories:A Weakly-supervised Approach

3D shapes captured by scanning devices are often incomplete due to occlusion. 3D shape completion methods have been explored to tackle this limitation. However, most of these methods are only trained and tested on a subset of categories, resulting in poor generalization to unseen categories. In this paper, we introduce a novel weakly-supervised framework to reconstruct the complete shapes from unseen categories. We first propose an end-to-end prior-assisted shape learning network that leverages data from the seen categories to infer a coarse shape. Specifically, we construct a prior bank consisting of representative shapes from the seen categories. Then, we design a multi-scale pattern correlation module for learning the complete shape of the input by analyzing the correlation between local patterns within the input and the priors at various scales. In addition, we propose a self-supervised shape refinement model to further refine the coarse shape. Considering the shape variability of 3D objects across categories, we construct a category-specific prior bank to facilitate shape refinement. Then, we devise a voxel-based partial matching loss and leverage the partial scans to drive the refinement process. Extensive experimental results show that our approach is superior to state-of-the-art methods by a large margin.

cs.CV↗

arXiv2024

SpatialCodec: Neural Spatial Speech Coding

In this work, we address the challenge of encoding speech captured by a microphone array using deep learning techniques with the aim of preserving and accurately reconstructing crucial spatial cues embedded in multi-channel recordings. We propose a neural spatial audio coding framework that achieves a high compression ratio, leveraging single-channel neural sub-band codec and SpatialCodec. Our approach encompasses two phases: (i) a neural sub-band codec is designed to encode the reference channel with low bit rates, and (ii), a SpatialCodec captures relative spatial information for accurate multi-channel reconstruction at the decoder end. In addition, we also propose novel evaluation metrics to assess the spatial cue preservation: (i) spatial similarity, which calculates cosine similarity on a spatially intuitive beamspace, and (ii), beamformed audio quality. Our system shows superior spatial performance compared with high bitrate baselines and black-box neural architecture. Demos are available at https://xzwy.github.io/SpatialCodecDemo. Codes and models are available at https://github.com/XZWY/SpatialCodec.

cs.SD↗

arXiv2024

Call Me When Necessary: LLMs can Efficiently and Faithfully Reason over Structured Environments

Large Language Models (LLMs) have shown potential in reasoning over structured environments, e.g., knowledge graph and table. Such tasks typically require multi-hop reasoning, i.e., match natural language utterance with instances in the environment. Previous methods leverage LLMs to incrementally build a reasoning path, where the LLMs either invoke tools or pick up schemas by step-by-step interacting with the environment. We propose Reasoning-Path-Editing (Readi), a novel framework where LLMs can efficiently and faithfully reason over structured environments. In Readi, LLMs initially generate a reasoning path given a query, and edit the path only when necessary. We instantiate the path on structured environments and provide feedback to edit the path if anything goes wrong. Experimental results on three KGQA and two TableQA datasets show the effectiveness of Readi, significantly surpassing previous LLM-based methods (by 9.1% Hit@1 on WebQSP, 12.4% on MQA-3H and 9.5% on WTQ), comparable with state-of-the-art fine-tuned methods (67% on CWQ and 74.7% on WebQSP) and substantially boosting the vanilla LLMs (by 14.9% on CWQ). Our code will be available on https://aka.ms/readi.

cs.CL↗

arXiv2024

Improving density matrix electronic structure method by deep learning

The combination of deep learning and ab initio materials calculations is emerging as a trending frontier of materials science research, with deep-learning density functional theory (DFT) electronic structure being particularly promising. In this work, we introduce a neural-network method for modeling the DFT density matrix, a fundamental yet previously unexplored quantity in deep-learning electronic structure. Utilizing an advanced neural network framework that leverages the nearsightedness and equivariance properties of the density matrix, the method demonstrates high accuracy and excellent generalizability in multiple example studies, as well as capability to precisely predict charge density and reproduce other electronic structure properties. Given the pivotal role of the density matrix in DFT as well as other computational methods, the current research introduces a novel approach to the deep-learning study of electronic structure properties, opening up new opportunities for deep-learning enhanced computational materials study.

physics.comp-ph↗

arXiv2024

Text-Queried Target Sound Event Localization

Sound event localization and detection (SELD) aims to determine the appearance of sound classes, together with their Direction of Arrival (DOA). However, current SELD systems can only predict the activities of specific classes, for example, 13 classes in DCASE challenges. In this paper, we propose text-queried target sound event localization (SEL), a new paradigm that allows the user to input the text to describe the sound event, and the SEL model can predict the location of the related sound event. The proposed task presents a more user-friendly way for human-computer interaction. We provide a benchmark study for the proposed task and perform experiments on datasets created by simulated room impulse response (RIR) and real RIR to validate the effectiveness of the proposed methods. We hope that our benchmark will inspire the interest and additional research for text-queried sound source localization.

eess.AS↗

arXiv2024

TaskWeaver: A Code-First Agent Framework

Large Language Models (LLMs) have shown impressive abilities in natural language understanding and generation, leading to their widespread use in applications such as chatbots and virtual assistants. However, existing LLM frameworks face limitations in handling domain-specific data analytics tasks with rich data structures. Moreover, they struggle with flexibility to meet diverse user requirements. To address these issues, TaskWeaver is proposed as a code-first framework for building LLM-powered autonomous agents. It converts user requests into executable code and treats user-defined plugins as callable functions. TaskWeaver provides support for rich data structures, flexible plugin usage, and dynamic plugin selection, and leverages LLM coding capabilities for complex logic. It also incorporates domain-specific knowledge through examples and ensures the secure execution of generated code. TaskWeaver offers a powerful and flexible framework for creating intelligent conversational agents that can handle complex tasks and adapt to domain-specific scenarios. The code is open sourced at https://github.com/microsoft/TaskWeaver/.

cs.AI↗

arXiv2024

Multi-Channel Multi-Speaker ASR Using Target Speaker's Solo Segment

In the field of multi-channel, multi-speaker Automatic Speech Recognition (ASR), the task of discerning and accurately transcribing a target speaker's speech within background noise remains a formidable challenge. Traditional approaches often rely on microphone array configurations and the information of the target speaker's location or voiceprint. This study introduces the Solo Spatial Feature (Solo-SF), an innovative method that utilizes a target speaker's isolated speech segment to enhance ASR performance, thereby circumventing the need for conventional inputs like microphone array layouts. We explore effective strategies for selecting optimal solo segments, a crucial aspect for Solo-SF's success. Through evaluations conducted on the AliMeeting dataset and AISHELL-1 simulations, Solo-SF demonstrates superior performance over existing techniques, significantly lowering Character Error Rates (CER) in various test conditions. Our findings highlight Solo-SF's potential as an effective solution for addressing the complexities of multi-channel, multi-speaker ASR tasks.

eess.AS↗