arXiv subjects

Yuehai Wang

Publications and source records attributed to Yuehai Wang.

18 recordsLinked to original sources

arXiv2026

Photonic-Implemented Efficient Deep Quantum Neural Network via Virtual-Driven Hilbert Space Expansion

The growing computational demands of classical neural networks have intensified the search for energy-efficient and powerful computational alternatives. Quantum neural networks (QNNs) implemented on integrated photonic platforms offer a compelling avenue, offering exceptional computational power enhancements, with inherent programmability and scalability of integrated architectures. A critical challenge, however, is implementing the fundamental non-unitary and nonlinear activation function of QNNs within a linear quantum photonic system. Existing strategies, such as the adding ancillary qubits and measurement-based feedback or forward are constrained by high qubit resource costs, overhead devices, and poor cascadability. Here, we propose a novel deep photonic QNN with an expanded computational Hilbert space via input replication and mode expansion, which enables the realization of effective non-unitary and nonlinear activation on a linear programmable quantum photonic chip. This approach eliminates the need for physical ancillary qubits, measurement-induced qubit consumption and the measurement device burden, thereby significantly reduce resource costs. The fabricated chip integrates four high-quality entanglement sources and a programmable high-dimensional interferometric network, enabling a two-hidden-layer QNN that exhibits dimension-enhanced expressivity over the existing QNN architectures. We demonstrate its capabilities across diverse tasks, including nonlinear classification, image generation, and quantum Gibbs state preparation. This work establishes a scalable and efficient architecture toward practical quantum deep learning systems capable of tackling problems beyond the reach of classical computation.

quant-ph↗

arXiv2025

Empathy Omni: Enabling Empathetic Speech Response Generation through Large Language Models

With the development of speech large language models (speech LLMs), users can now interact directly with assistants via speech. However, most existing models only convert response content into speech without fully capturing the rich emotional cues in user queries, where the same sentence may convey different meanings depending on the expression. Emotional understanding is thus essential for improving human-machine interaction. Most empathetic speech LLMs rely on massive datasets, demanding high computational cost. A key challenge is to build models that generate empathetic responses with limited data and without large-scale training. To this end, we propose Emotion Omni, a model that understands emotional content in user speech and generates empathetic responses. We further developed a data pipeline to construct a 200k emotional dialogue dataset supporting empathetic speech assistants. Experiments show that Emotion Omni achieves comparable instruction-following ability without large-scale pretraining, while surpassing existing models in speech quality (UTMOS:4.41) and empathy (Emotion GPT Score: 3.97). These results confirm its improvements in both speech fidelity and emotional expressiveness. Demos are available at https://w311411.github.io/omni_demo/.

cs.CL↗

arXiv2025

HunyuanVideo-Foley: Multimodal Diffusion with Representation Alignment for High-Fidelity Foley Audio Generation

Recent advances in video generation produce visually realistic content, yet the absence of synchronized audio severely compromises immersion. To address key challenges in video-to-audio generation, including multimodal data scarcity, modality imbalance and limited audio quality in existing methods, we propose HunyuanVideo-Foley, an end-to-end text-video-to-audio framework that synthesizes high-fidelity audio precisely aligned with visual dynamics and semantic context. Our approach incorporates three core innovations: (1) a scalable data pipeline curating 100k-hour multimodal datasets through automated annotation; (2) a representation alignment strategy using self-supervised audio features to guide latent diffusion training, efficiently improving audio quality and generation stability; (3) a novel multimodal diffusion transformer resolving modal competition, containing dual-stream audio-video fusion through joint attention, and textual semantic injection via cross-attention. Comprehensive evaluations demonstrate that HunyuanVideo-Foley achieves new state-of-the-art performance across audio fidelity, visual-semantic alignment, temporal alignment and distribution matching. The demo page is available at: https://szczesnys.github.io/hunyuanvideo-foley/.

eess.AS↗

arXiv2025

A Large Language Model Powered Integrated Circuit Footprint Geometry Understanding

Printed-Circuit-board (PCB) footprint geometry labeling of integrated circuits (IC) is essential in defining the physical interface between components and the PCB layout, requiring exceptional visual perception proficiency. However, due to the unstructured footprint drawing and abstract diagram annotations, automated parsing and accurate footprint geometry modeling remain highly challenging. Despite its importance, no methods currently exist for automated package geometry labeling directly from IC mechanical drawings. In this paper, we first investigate the visual perception performance of Large Multimodal Models (LMMs) when solving IC footprint geometry understanding. Our findings reveal that current LMMs severely suffer from inaccurate geometric perception, which hinders their performance in solving the footprint geometry labeling problem. To address these limitations, we propose LLM4-IC8K, a novel framework that treats IC mechanical drawings as images and leverages LLMs for structured geometric interpretation. To mimic the step-by-step reasoning approach used by human engineers, LLM4-IC8K addresses three sub-tasks: perceiving the number of pins, computing the center coordinates of each pin, and estimating the dimensions of individual pins. We present a two-stage framework that first trains LMMs on synthetically generated IC footprint diagrams to learn fundamental geometric reasoning and then fine-tunes them on real-world datasheet drawings to enhance robustness and accuracy in practical scenarios. To support this, we introduce ICGeo8K, a multi-modal dataset with 8,608 labeled samples, including 4138 hand-crafted IC footprint samples and 4470 synthetically generated samples. Extensive experiments demonstrate that our model outperforms state-of-the-art LMMs on the proposed benchmark.

cs.CV↗

arXiv2025

Programmable time-frequency mode encoded quantum state generator for silicon-on-insulator platform

We propose a method for the programmable generation of time-frequency mode (TFM) encoded quantum states of light on the silicon-on-insulator (SOI) platform. The state generator consists of an N-tap finite impulse response filter and a Mach-Zehnder interferometer (MZI)-based coupled-ring resonator. Through numerical simulations, its capability of producing TFM-encoded maximally entangled states in two, three, and four dimensions is theoretically demonstrated, with fidelities of 0.950, 0.954, and 0.971, respectively.

quant-ph↗

arXiv2025

Ensemble-based Deep Multilayer Community Search

Multilayer graphs, consisting of multiple interconnected layers, are widely used to model diverse relationships in the real world. A community is a cohesive subgraph that offers valuable insights for analyzing (multilayer) graphs. Recently, there has been an emerging trend focused on searching query-driven communities within the multilayer graphs. However, existing methods for multilayer community search are either 1) rule-based, which suffer from structure inflexibility; or 2) learning-based, which rely on labeled data or fail to capture layer-specific characteristics. To address these, we propose EnMCS, an Ensemble-based unsupervised (i.e., label-free) Multilayer Community Search framework. EnMCS contains two key components, i.e., HoloSearch which identifies potential communities in each layer while integrating both layer-shared and layer-specific information, and EMerge which is an Expectation-Maximization (EM)-based method that synthesizes the potential communities from each layer into a consensus community. Specifically, HoloSearch first employs a graph-diffusion-based model that integrates three label-free loss functions to learn layer-specific and layer-shared representations for each node. Communities in each layer are then identified based on nodes that exhibit high similarity in layer-shared representations while demonstrating low similarity in layer-specific representations w.r.t. the query nodes. To account for the varying layer-specific characteristics of each layer when merging communities, EMerge models the error rates of layers and true community as latent variables. It then employs the EM algorithm to simultaneously minimize the error rates of layers and predict the final consensus community through iterative maximum likelihood estimation. Experiments over 10 real-world datasets highlight the superiority of EnMCS in terms of both efficiency and effectiveness.

cs.SI↗

arXiv2024

Engineering biphoton spectral wavefunction in a silicon micro-ring resonator with split resonances

Frequency-time is a degree of freedom suitable for photonic high-dimensional entanglement, with advantages such as compatibility with single-mode devices and insensitivity to dispersion. The engineering control of the frequency-time amplitude of a photon's electric field has been demonstrated on platforms with second-order optical nonlinearity. For integrated photonic platforms with only third-order optical nonlinearity, the engineered generation of the state remains unexplored. Here, we demonstrate a cavity-enhanced photon-pair source on the silicon-on-insulator (SOI) platform that can generate both separable states and controllable entangled states in the frequency domain without post-manipulation. By choosing different resonance combinations and employing on-chip optical field differentiation, we achieve independent control over two functions that affect the joint spectral intensity (JSI) of the state. A semi-analytical model is derived to simulate the biphoton spectral wavefunction in the presence of resonance splitting and pump differentiation, and its parameters can be fully determined through fitting-based parameter extraction from the resonator's measured linear response. The measured spectral purity for the separable state is $95.5\pm 1.2\%$, while the measured JSIs for the entangled states show two- or four-peaked functions in two-dimensional frequency space. The experiments and simulations demonstrate the capacity to manipulate the frequency-domain wavefunction in a silicon-based device, which is promising for applications like quantum information processing using pulsed temporal-mode encoding or long-distance quantum key distribution.

quant-ph↗

arXiv2024

FLY-TTS: Fast, Lightweight and High-Quality End-to-End Text-to-Speech Synthesis

While recent advances in Text-To-Speech synthesis have yielded remarkable improvements in generating high-quality speech, research on lightweight and fast models is limited. This paper introduces FLY-TTS, a new fast, lightweight and high-quality speech synthesis system based on VITS. Specifically, 1) We replace the decoder with ConvNeXt blocks that generate Fourier spectral coefficients followed by the inverse short-time Fourier transform to synthesize waveforms; 2) To compress the model size, we introduce the grouped parameter-sharing mechanism to the text encoder and flow-based model; 3) We further employ the large pre-trained WavLM model for adversarial training to improve synthesis quality. Experimental results show that our model achieves a real-time factor of 0.0139 on an Intel Core i9 CPU, 8.8x faster than the baseline (0.1221), with a 1.6x parameter compression. Objective and subjective evaluations indicate that FLY-TTS exhibits comparable speech quality to the strong baseline.

eess.AS↗

arXiv2024

Quantum Generative Adversarial Networks in a Silicon Photonic Chip with Maximum Expressibility

Generative adversarial networks (GANs) have achieved remarkable success with realistic tasks such as creating realistic images, texts, and audio. Combining GANs and quantum computing, quantum GANs are thought to have an exponential advantage over their classical counterparts due to the stronger expressibility of quantum circuits. In this research, a two-qubit silicon quantum photonic chip is created, capable of executing arbitrary controlled-unitary (CU) operations and generating any 2-qubit pure state, thus making it an excellent platform for quantum GANs. To capture complex data patterns, a hybrid generator is proposed to inject nonlinearity into quantum GANs. As a demonstration, three generative tasks, covering both pure quantum versions of GANs (PQ-GAN) and hybrid quantum-classical GANs (HQC-GANs), are successfully carried out on the chip, including high-fidelity single-qubit state learning, classical distributions loading, and compressed image production. The experiment results prove that silicon quantum photonic chips have great potential in generative learning applications.

quant-ph↗

arXiv2024

Continuous Target Speech Extraction: Enhancing Personalized Diarization and Extraction on Complex Recordings

Target speaker extraction (TSE) aims to extract the target speaker's voice from the input mixture. Previous studies have concentrated on high-overlapping scenarios. However, real-world applications usually meet more complex scenarios like variable speaker overlapping and target speaker absence. In this paper, we introduces a framework to perform continuous TSE (C-TSE), comprising a target speaker voice activation detection (TSVAD) and a TSE model. This framework significantly improves TSE performance on similar speakers and enhances personalization, which is lacking in traditional diarization methods. In detail, unlike conventional TSVAD deployed to refine the diarization results, the proposed Attention-target speaker voice activation detection (A-TSVAD) directly generates timestamps of the target speaker. We also explore some different integration methods of A-TSVAD and TSE by comparing the cascaded and parallel methods. The framework's effectiveness is assessed using a range of metrics, including diarization and enhancement metrics. Our experiments demonstrate that A-TSVAD outperforms conventional methods in reducing diarization errors. Furthermore, the integration of A-TSVAD and TSE in a sequential cascaded manner further enhances extraction accuracy.

cs.SD↗

arXiv2024

Audio Deepfake Detection with Self-Supervised WavLM and Multi-Fusion Attentive Classifier

With the rapid development of speech synthesis and voice conversion technologies, Audio Deepfake has become a serious threat to the Automatic Speaker Verification (ASV) system. Numerous countermeasures are proposed to detect this type of attack. In this paper, we report our efforts to combine the self-supervised WavLM model and Multi-Fusion Attentive classifier for audio deepfake detection. Our method exploits the WavLM model to extract features that are more conducive to spoofing detection for the first time. Then, we propose a novel Multi-Fusion Attentive (MFA) classifier based on the Attentive Statistics Pooling (ASP) layer. The MFA captures the complementary information of audio features at both time and layer levels. Experiments demonstrate that our methods achieve state-of-the-art results on the ASVspoof 2021 DF set and provide competitive results on the ASVspoof 2019 and 2021 LA set.

eess.AS↗

arXiv2022

Masked Acoustic Unit for Mispronunciation Detection and Correction

Computer-Assisted Pronunciation Training (CAPT) plays an important role in language learning. Conventional ASR-based CAPT methods require expensive annotation of the ground truth pronunciation for the supervised training. Meanwhile, certain undefined non-native phonemes cannot be correctly classified into standard phonemes, making the annotation process challenging and subjective. On the other hand, ASR-based CAPT methods only give the learner text-based feedback about the mispronunciation, but cannot teach the learner how to pronounce the sentence correctly. To solve these limitations, we propose to use the acoustic unit (AU) as the intermediary feature for both mispronunciation detection and correction. The proposed method uses the masked AU sequence and the target phonemes to detect the error AU and then corrects it. This method can give the learner speech-based self-imitating feedback, making our CAPT powerful for education.

eess.AS↗

arXiv2021

Accent Recognition with Hybrid Phonetic Features

The performance of voice-controlled systems is usually influenced by accented speech. To make these systems more robust, the frontend accent recognition (AR) technologies have received increased attention in recent years. As accent is a high-level abstract feature that has a profound relationship with the language knowledge, AR is more challenging than other language-agnostic audio classification tasks. In this paper, we use an auxiliary automatic speech recognition (ASR) task to extract language-related phonetic features. Furthermore, we propose a hybrid structure that incorporates the embeddings of both a fixed acoustic model and a trainable acoustic model, making the language-related acoustic feature more robust. We conduct several experiments on the Accented English Speech Recognition Challenge (AESRC) 2020 dataset. The results demonstrate that our approach can obtain a 6.57% relative improvement on the validation set. We also get a 7.28% relative improvement on the final test set for this competition, showing the merits of the proposed method.

eess.AS↗

arXiv2021

Text-Conditioned Transformer for Automatic Pronunciation Error Detection

Automatic pronunciation error detection (APED) plays an important role in the domain of language learning. As for the previous ASR-based APED methods, the decoded results need to be aligned with the target text so that the errors can be found out. However, since the decoding process and the alignment process are independent, the prior knowledge about the target text is not fully utilized. In this paper, we propose to use the target text as an extra condition for the Transformer backbone to handle the APED task. The proposed method can output the error states with consideration of the relationship between the input speech and the target text in a fully end-to-end fashion.Meanwhile, as the prior target text is used as a condition for the decoder input, the Transformer works in a feed-forward manner instead of autoregressive in the inference stage, which can significantly boost the speed in the actual deployment. We set the ASR-based Transformer as the baseline APED model and conduct several experiments on the L2-Arctic dataset. The results demonstrate that our approach can obtain 8.4\% relative improvement on the $F_1$ score metric.

eess.AS↗

arXiv2020

Collaborative Distillation for Ultra-Resolution Universal Style Transfer

Universal style transfer methods typically leverage rich representations from deep Convolutional Neural Network (CNN) models (e.g., VGG-19) pre-trained on large collections of images. Despite the effectiveness, its application is heavily constrained by the large model size to handle ultra-resolution images given limited memory. In this work, we present a new knowledge distillation method (named Collaborative Distillation) for encoder-decoder based neural style transfer to reduce the convolutional filters. The main idea is underpinned by a finding that the encoder-decoder pairs construct an exclusive collaborative relationship, which is regarded as a new kind of knowledge for style transfer models. Moreover, to overcome the feature size mismatch when applying collaborative distillation, a linear embedding loss is introduced to drive the student network to learn a linear embedding of the teacher's features. Extensive experiments show the effectiveness of our method when applied to different universal style transfer approaches (WCT and AdaIN), even if the model size is reduced by 15.5 times. Especially, on WCT with the compressed models, we achieve ultra-resolution (over 40 megapixels) universal style transfer on a 12GB GPU for the first time. Further experiments on optimization-based stylization scheme show the generality of our algorithm on different stylization paradigms. Our code and trained models are available at https://github.com/mingsun-tse/collaborative-distillation.

cs.CV↗

arXiv2019

Structured Pruning for Efficient ConvNets via Incremental Regularization

Parameter pruning is a promising approach for CNN compression and acceleration by eliminating redundant model parameters with tolerable performance degrade. Despite its effectiveness, existing regularization-based parameter pruning methods usually drive weights towards zero with large and constant regularization factors, which neglects the fragility of the expressiveness of CNNs, and thus calls for a more gentle regularization scheme so that the networks can adapt during pruning. To achieve this, we propose a new and novel regularization-based pruning method, named IncReg, to incrementally assign different regularization factors to different weights based on their relative importance. Empirical analysis on CIFAR-10 dataset verifies the merits of IncReg. Further extensive experiments with popular CNNs on CIFAR-10 and ImageNet datasets show that IncReg achieves comparable to even better results compared with state-of-the-arts. Our source codes and trained models are available here: https://github.com/mingsun-tse/caffe_increg.

cs.LG↗

arXiv2018

Structured Pruning for Efficient ConvNets via Incremental Regularization

Parameter pruning is a promising approach for CNN compression and acceleration by eliminating redundant model parameters with tolerable performance loss. Despite its effectiveness, existing regularization-based parameter pruning methods usually drive weights towards zero with large and constant regularization factors, which neglects the fact that the expressiveness of CNNs is fragile and needs a more gentle way of regularization for the networks to adapt during pruning. To solve this problem, we propose a new regularization-based pruning method (named IncReg) to incrementally assign different regularization factors to different weight groups based on their relative importance, whose effectiveness is proved on popular CNNs compared with state-of-the-art methods.

cs.CV↗

arXiv2018

Structured Probabilistic Pruning for Convolutional Neural Network Acceleration

In this paper, we propose a novel progressive parameter pruning method for Convolutional Neural Network acceleration, named Structured Probabilistic Pruning (SPP), which effectively prunes weights of convolutional layers in a probabilistic manner. Unlike existing deterministic pruning approaches, where unimportant weights are permanently eliminated, SPP introduces a pruning probability for each weight, and pruning is guided by sampling from the pruning probabilities. A mechanism is designed to increase and decrease pruning probabilities based on importance criteria in the training process. Experiments show that, with 4x speedup, SPP can accelerate AlexNet with only 0.3% loss of top-5 accuracy and VGG-16 with 0.8% loss of top-5 accuracy in ImageNet classification. Moreover, SPP can be directly applied to accelerate multi-branch CNN networks, such as ResNet, without specific adaptations. Our 2x speedup ResNet-50 only suffers 0.8% loss of top-5 accuracy on ImageNet. We further show the effectiveness of SPP on transfer learning tasks.

cs.LG↗