arXiv subjects

Zhenyu Zhou

Publications and source records attributed to Zhenyu Zhou.

At least 19 recordsLinked to original sources

arXiv2026

DICE: Distilling Classifier-Free Guidance into Text Embeddings

Text-to-image diffusion models are capable of generating high-quality images, but suboptimal pre-trained text representations often result in these images failing to align closely with the given text prompts. Classifier-free guidance (CFG) is a popular and effective technique for improving text-image alignment in the generative process. However, CFG introduces significant computational overhead. In this paper, we present DIstilling CFG by sharpening text Embeddings (DICE) that replaces CFG in the sampling process with half the computational complexity while maintaining similar generation quality. DICE distills a CFG-based text-to-image diffusion model into a CFG-free version by refining text embeddings to replicate CFG-based directions. In this way, we avoid the computational drawbacks of CFG, enabling high-quality, well-aligned image generation at a fast sampling speed. Furthermore, examining the enhancement pattern, we identify the underlying mechanism of DICE that sharpens specific components of text embeddings to preserve semantic information while enhancing fine-grained details. Extensive experiments on multiple Stable Diffusion v1.5 variants, SDXL, and PixArt-$α$ demonstrate the effectiveness of our method. Code is available at https://github.com/zju-pi/dice.

cs.CV↗

arXiv2026

Analyzing and Improving Fast Sampling of Text-to-Image Diffusion Models

Text-to-image diffusion models have achieved unprecedented success but still struggle to produce high-quality results under limited sampling budgets. Existing training-free sampling acceleration methods are typically developed independently, leaving the overall performance and compatibility among these methods unexplored. In this paper, we bridge this gap by systematically elucidating the design space, and our comprehensive experiments identify the sampling time schedule as the most pivotal factor. Inspired by the geometric properties of diffusion models revealed through the Frenet-Serret formulas, we propose constant total rotation schedule (TORS), a scheduling strategy that ensures uniform geometric variation along the sampling trajectory. TORS outperforms previous training-free acceleration methods and produces high-quality images with 10 sampling steps on Flux.1-Dev and Stable Diffusion 3.5. Extensive experiments underscore the adaptability of our method to unseen models, hyperparameters, and downstream applications. Code is available at https://github.com/zju-pi/TORS.

cs.CV↗

arXiv2026

Optimized finite-$β$ tokamak-stellarator hybrid configurations achieved by planar dipole-field coils

Tokamak--stellarator hybrids seek to combine tokamak-like compactness and confinement with stellarator-like externally generated rotational transform and steady-state operation. In this work, we build on the recent tokamak--stellarator hybrid study using planar dipole-field coils (PDCs) [Yu et al., arXiv:2605.03599], in which the fixed-position, programmable coils on an axisymmetric winding surface generate flexible three-dimensional shaping fields. Using single-stage free-boundary optimization of coil currents and plasma-equilibrium parameters, we construct vacuum and finite-$β$ configurations. The vacuum cases show controllable external transform and magnetic well. The finite-$β$ cases accommodate various density, temperature, and pressure profiles, producing quasi-axisymmetric (QA) equilibria with self-consistent bootstrap current, favorable Mercier stability, and reduced demand for external current drive. Re-optimization enables $β$ ramp-up and access to different field-period QA branches with moderate coil-current changes. At large rotational transform, a toroidally omnigenous (TO)-like configuration exhibits more favorable infinite-$n$ ideal-ballooning behavior than a QA reference with matched profiles, even though ballooning stability is not directly optimized for. These results demonstrate that PDCs provide a flexible platform for achieving optimized finite-$β$ hybrid configurations.

physics.plasm-ph↗

arXiv2026

Integrated magnonic neural circuits based on nonlinear wave neurons

Artificial intelligence is driving intense interest in alternative computing hardware capable of neural information processing beyond conventional charge-based electronics. Among emerging approaches, wave-based computing promises highly parallel and energy-efficient operation, but scalable physical neural hardware has remained elusive because wave systems generally lack cascadable nonlinear neurons with signal regeneration and phase-robust operation. Here we demonstrate integrated magnonic neural circuits based on nonlinear threshold neurons realized in nanoscale yttrium iron garnet waveguides. The neurons perform weighted summation of multiple spin-wave inputs, while a pump-controlled nonlinear activation defines continuously tunable firing thresholds. Owing to deeply nonlinear spin-wave dynamics, the activated neurons emit self-normalized outputs whose intensities are largely independent of the input amplitudes, while nonlinear phase self-adjustment suppresses sensitivity to the relative input phases, enabling deterministic neuron-to-neuron cascading without external signal restoration. We experimentally realize programmable threshold neurons, reconfigurable weighted classification and deterministic cascading between sequential neuronal stages, and further demonstrate reconfigurable physical pattern recognition in a seven-neuron integrated magnonic circuit through experimental classification of the binary letter patterns 'HUST'. These results establish nonlinear magnons as a scalable platform for integrated neural hardware and position nonlinear wave dynamics as a general paradigm for physical neuromorphic computing.

cond-mat.mtrl-sci↗

arXiv2026

Subspace-Guided Semantic and Topological Invariant Registration for Annotation-Free Ultrasound Plane Quality Control

Reliable quality control (QC) of ultrasound images is essential for both real-time acquisition guidance and retrospective clinical audit, yet existing approaches rely heavily on per-plane annotations, or employ pseudo-labeling prone to systematic bias under spatial deformations inherent in clinical acquisition. We present STRIQ, a registration-driven framework that recasts annotation-free US plane quality control as a subspace-guided consistency measurement problem. Specifically, STRIQ introduces a Latent Registration Aligner (LRA) to establish hierarchical feature space correspondences between query images and variance-driven anchors, which are autonomously distilled from unlabeled data via a variance spectrum criterion to serve as structurally stable prototypes. To further disambiguate anatomical planes and mitigate negative knowledge transfer, we propose an Orthogonal Knowledge Subspace (OKS) module. The OKS decomposes plane-specific representations into mutually orthogonal subspaces, enabling fine-grained expert collaboration while preventing inter-plane interference, ensuring that the quality metric is grounded in principled subspace proximity. Extensive experiments on the in-house US4QA and public CAMUS datasets demonstrate that STRIQ achieves state-of-the-art correlation with clinical quality scores, establishing a new paradigm for annotation-free, real-time reliable ultrasound quality control. Our code is available at https://github.com/zhcz328/STRIQ.

cs.CV↗

arXiv2026

True random number generation through stochastic magnonic bistability

True random number generators (TRNGs) underpin modern cryptography, yet existing implementations face fundamental trade-offs between speed, scalability, and entropy quality. Here, we demonstrate that stochastic switching in the bistable regime of spin-wave dynamics provides a physical entropy source for high-quality random number generation. Our magnonic random number generator (mRNG), based on a lithography-patterned microstrip on yttrium iron garnet (YIG), exploits thermal fluctuations near the nonlinear bistable regime to generate random bitstreams that pass all 15 NIST SP 800-22 statistical tests at rates with 20 Mb/s. We implement a random-bit multiplier using synchronized mRNG units and demonstrate scalability to 200-nm-wide nanoscale waveguides, establishing spin-wave bistability as a viable physical entropy source for integrated random number generation.

cond-mat.mtrl-sci↗

arXiv2026

Integrated magnonic chip using cascaded logic

The transistor transformed not only electronics but everyday life, and the integrated circuit - now simply the "chip" - made computation scalable and ubiquitous. Magnonics has long promised a parallel path to low-energy information processing by using spin waves instead of charge. Progress, however, has been limited by two fundamental obstacles: intrinsic attenuation of spin waves and the requirement for precisely normalised output intensity and input phase to ensure reliable logic operation - conditions that are difficult to maintain in large-scale circuits owing to inevitable imperfections. Here, we report an integrated magnonic circuit that overcomes both limitations through engineered nonlinearity in nanoscale yttrium iron garnet waveguides. Nonlinear self-adjustment of the spin wave phase renders logic operation insensitive to the relative phases of the inputs, while a deeply nonlinear, threshold-activated self-normalised excitation restores and standardises the output intensity. Using space-resolved micro-focused Brillouin light scattering, we demonstrate reconfigurable AND, OR and three-input majority gates and realise deterministic cascading across sequential stages, establishing a scalable on-chip logic primitive. The architecture operates with gigahertz frequencies, supports dynamic threshold control for functional reconfiguration, and is compatible with scalable integration, making it attractive for adaptive and neuromorphic computing. By resolving phase-independent operation and signal restoration at the level of device physics, this work advances magnonics from isolated proof-of-concept devices towards integrated magnonic chips that can complement advanced CMOS in energy-constrained computing tasks.

cond-mat.mtrl-sci↗

arXiv2025

Geometric Regularity in Deterministic Sampling Dynamics of Diffusion-based Generative Models

Diffusion-based generative models employ stochastic differential equations (SDEs) and their equivalent probability flow ordinary differential equations (ODEs) to establish a smooth transformation between complex high-dimensional data distributions and tractable prior distributions. In this paper, we reveal a striking geometric regularity in the deterministic sampling dynamics of diffusion generative models: each simulated sampling trajectory along the gradient field lies within an extremely low-dimensional subspace, and all trajectories exhibit an almost identical boomerang shape, regardless of the model architecture, applied conditions, or generated content. We characterize several intriguing properties of these trajectories, particularly under closed-form solutions based on kernel-estimated data modeling. We also demonstrate a practical application of the discovered trajectory regularity by proposing a dynamic programming-based scheme to better align the sampling time schedule with the underlying trajectory structure. This simple strategy requires minimal modification to existing deterministic numerical solvers, incurs negligible computational overhead, and achieves superior image generation performance, especially in regions with only 5 - 10 function evaluations.

cs.LG↗

arXiv2025

Scaling Up Data Parallelism in Decentralized Deep Learning

Although it has been extensively explored in theory, decentralized learning is not yet green-lighted for production use, largely due to a lack of stability, scalability, and generality in large scale DNN training. To shed light on the production use of decentralized learning, this work studies decentralized data parallel training at scale. To this end, we introduce a benchmarking framework, namely DBench, to host both centralized and decentralized DNN training. Building upon DBench, we introduce a benchmarking methodology to uncover the correlations between model accuracy and the variances of parameter tensors by varying communication graphs and training scales. Based on the benchmarking results, we observe that, (1) Similar to centralized learning, decentralized data parallel training also presents the issues of scalability and generality when the training scales up; (2) The model accuracy of decentralized learning is correlated to the number of connections in a communication graph; (3) The model accuracy of decentralized learning is surprisingly sensitive to the variance of parameter tensors across model replicas. Built upon the observations, we propose Ada, a decentralized adaptive approach that performs large scale DNN training following a decentralized SGD method and adapting the communication graph in use dynamically throughout training iterations. We apply Ada on large scale training and observe that Ada can obtain the best convergence rates consistently in decentralized DNN training, and delivers equally or comparably good model accuracy for all sample applications as centralized learning does, even when training ResNet50 for ImageNet-1K on the scale of 1008 GPUs.

cs.LG↗

arXiv2025

Neural Scoring: A Refreshed End-to-End Approach for Speaker Recognition in Complex Conditions

Modern speaker verification systems primarily rely on speaker embeddings, followed by verification based on cosine similarity between the embedding vectors of the enrollment and test utterances. While effective, these methods struggle with multi-talker speech due to the unidentifiability of embedding vectors. In this paper, we propose Neural Scoring (NS), a refreshed end-to-end framework that directly estimates verification posterior probabilities without relying on test-side embeddings, making it more robust to complex conditions, e.g., with multiple talkers. To make the training of such an end-to-end model more efficient, we introduce a large-scale trial e2e training (LtE2E) strategy, where each test utterance pairs with a set of enrolled speakers, thus enabling the processing of large-scale verification trials per batch. Experiments on the VoxCeleb dataset demonstrate that NS consistently outperforms both the baseline and competitive methods across various conditions, achieving an overall 70.36% reduction in EER compared to the baseline.

cs.SD↗

arXiv2025

An Investigation on Speaker Augmentation for End-to-End Speaker Extraction

Target confusion, defined as occasional switching to non-target speakers, poses a key challenge for end-to-end speaker extraction (E2E-SE) systems. We argue that this problem is largely caused by the lack of generalizability and discrimination of the speaker embeddings, and introduce a simple yet effective speaker augmentation strategy to tackle the problem. Specifically, we propose a time-domain resampling and rescaling pipeline that alters speaker traits while preserving other speech properties. This generates a variety of pseudo-speakers to help establish a generalizable speaker embedding space, while the speaker-trait-specific augmentation creates hard samples that force the model to focus on genuine speaker characteristics. Experiments on WSJ0-2Mix and LibriMix show that our method mitigates the target confusion and improves extraction performance. Moreover, it can be combined with metric learning, another effective approach to address target confusion, leading to further gains.

cs.SD↗

arXiv2024

Improving the Reliability of Cable Broadband Networks via Proactive Network Maintenance

Cable broadband networks are one of the few "last-mile" broadband technologies widely available in the U.S. Unfortunately, they have poor reliability after decades of deployment. The cable industry proposed a framework called Proactive Network Maintenance (PNM) to diagnose the cable networks. However, there is little public knowledge or systematic study on how to use these data to detect and localize cable network problems. Existing tools in the public domain have prohibitive high false-positive rates. In this paper, we propose CableMon, the first public-domain system that applies machine learning techniques to PNM data to improve the reliability of cable broadband networks. CableMon tackles two key challenges faced by cable ISPs: accurately detecting failures, and distinguishing whether a failure occurs within a network or at a subscriber's premise. CableMon uses statistical models to generate features from time series data and uses customer trouble tickets as hints to infer abnormal/failure thresholds for these generated features. Further, CableMon employs an unsupervised learning model to group cable devices sharing similar anomalous patterns and effectively identify impairments that occur inside a cable network and impairments occur at a subscriber's premise, as these two different faults require different types of technical personnel to repair them. We use eight months of PNM data and customer trouble tickets from an ISP and experimental deployment to evaluate CableMon's performance. Our evaluation results show that CableMon can effectively detect and distinguish failures from PNM data and outperforms existing public-domain tools.

cs.NI↗

arXiv2024

TelApart: Differentiating Network Faults from Customer-Premise Faults in Cable Broadband Networks

Two types of radio frequency (RF) impairments frequently occur in a cable broadband network: impairments that occur inside a cable network and impairments occur at the edge of the broadband network, i.e., in a subscriber's premise. Differentiating these two types of faults is important, as different faults require different types of technical personnel to repair them. Presently, the cable industry lacks publicly available tools to automatically diagnose the type of fault. In this work, we present TelApart, a fault diagnosis system for cable broadband networks. TelApart uses telemetry data collected by the Proactive Network Maintenance (PNM) infrastructure in cable networks to effectively differentiate the type of fault. Integral to TelApart's design is an unsupervised machine learning model that groups cable devices sharing similar anomalous patterns together. We use metrics derived from an ISP's customer trouble tickets to programmatically tune the model's hyper-parameters so that an ISP can deploy TelApart in various conditions without hand-tuning its hyper-parameters. We also address the data challenge that the telemetry data collected by the PNM system contain numerous missing, duplicated, and unaligned data points. Using real-world data contributed by a cable ISP, we show that TelApart can effectively identify different types of faults.

cs.NI↗

arXiv2024

Towards More Accurate Fake Detection on Images Generated from Advanced Generative and Neural Rendering Models

The remarkable progress in neural-network-driven visual data generation, especially with neural rendering techniques like Neural Radiance Fields and 3D Gaussian splatting, offers a powerful alternative to GANs and diffusion models. These methods can produce high-fidelity images and lifelike avatars, highlighting the need for robust detection methods. In response, an unsupervised training technique is proposed that enables the model to extract comprehensive features from the Fourier spectrum magnitude, thereby overcoming the challenges of reconstructing the spectrum due to its centrosymmetric properties. By leveraging the spectral domain and dynamically combining it with spatial domain information, we create a robust multimodal detector that demonstrates superior generalization capabilities in identifying challenging synthetic images generated by the latest image synthesis techniques. To address the absence of a 3D neural rendering-based fake image database, we develop a comprehensive database that includes images generated by diverse neural rendering techniques, providing a robust foundation for evaluating and advancing detection methods.

cs.CV↗

arXiv2024

Simple and Fast Distillation of Diffusion Models

Diffusion-based generative models have demonstrated their powerful performance across various tasks, but this comes at a cost of the slow sampling speed. To achieve both efficient and high-quality synthesis, various distillation-based accelerated sampling methods have been developed recently. However, they generally require time-consuming fine tuning with elaborate designs to achieve satisfactory performance in a specific number of function evaluation (NFE), making them difficult to employ in practice. To address this issue, we propose Simple and Fast Distillation (SFD) of diffusion models, which simplifies the paradigm used in existing methods and largely shortens their fine-tuning time up to 1000$\times$. We begin with a vanilla distillation-based sampling method and boost its performance to state of the art by identifying and addressing several small yet vital factors affecting the synthesis efficiency and quality. Our method can also achieve sampling with variable NFEs using a single distilled model. Extensive experiments demonstrate that SFD strikes a good balance between the sample quality and fine-tuning costs in few-step image generation task. For example, SFD achieves 4.53 FID (NFE=2) on CIFAR-10 with only 0.64 hours of fine-tuning on a single NVIDIA A100 GPU. Our code is available at https://github.com/zju-pi/diff-sampler.

cs.CV↗

arXiv2024

Fast ODE-based Sampling for Diffusion Models in Around 5 Steps

Sampling from diffusion models can be treated as solving the corresponding ordinary differential equations (ODEs), with the aim of obtaining an accurate solution with as few number of function evaluations (NFE) as possible. Recently, various fast samplers utilizing higher-order ODE solvers have emerged and achieved better performance than the initial first-order one. However, these numerical methods inherently result in certain approximation errors, which significantly degrades sample quality with extremely small NFE (e.g., around 5). In contrast, based on the geometric observation that each sampling trajectory almost lies in a two-dimensional subspace embedded in the ambient space, we propose Approximate MEan-Direction Solver (AMED-Solver) that eliminates truncation errors by directly learning the mean direction for fast diffusion sampling. Besides, our method can be easily used as a plugin to further improve existing ODE-based samplers. Extensive experiments on image synthesis with the resolution ranging from 32 to 512 demonstrate the effectiveness of our method. With only 5 NFE, we achieve 6.61 FID on CIFAR-10, 10.74 FID on ImageNet 64$\times$64, and 13.20 FID on LSUN Bedroom. Our code is available at https://github.com/zju-pi/diff-sampler.

cs.CV↗

arXiv2024

A Geometric Perspective on Diffusion Models

Recent years have witnessed significant progress in developing effective training and fast sampling techniques for diffusion models. A remarkable advancement is the use of stochastic differential equations (SDEs) and their marginal-preserving ordinary differential equations (ODEs) to describe data perturbation and generative modeling in a unified framework. In this paper, we carefully inspect the ODE-based sampling of a popular variance-exploding SDE and reveal several intriguing structures of its sampling dynamics. We discover that the data distribution and the noise distribution are smoothly connected with a quasi-linear sampling trajectory and another implicit denoising trajectory that even converges faster. Meanwhile, the denoising trajectory governs the curvature of the corresponding sampling trajectory and its finite differences yield various second-order samplers used in practice. Furthermore, we establish a theoretical relationship between the optimal ODE-based sampling and the classic mean-shift (mode-seeking) algorithm, with which we can characterize the asymptotic behavior of diffusion models and identify the empirical score deviation. Code is available at \url{https://github.com/zju-pi/diff-sampler}.

cs.CV↗

arXiv2024

A Comprehensive Investigation on Speaker Augmentation for Speaker Recognition

Data augmentation (DA) has played a pivotal role in the success of deep speaker recognition. Current DA techniques primarily focus on speaker-preserving augmentation, which does not change the speaker trait of the speech and does not create new speakers. Recent research has shed light on the potential of speaker augmentation, which generates new speakers to enrich the training dataset. In this study, we delve into two speaker augmentation approaches: speed perturbation (SP) and vocal tract length perturbation (VTLP). Despite the empirical utilization of both methods, a comprehensive investigation into their efficacy is lacking. Our study, conducted using two public datasets, VoxCeleb and CN-Celeb, revealed that both SP and VTLP are proficient at generating new speakers, leading to significant performance improvements in speaker recognition. Furthermore, they exhibit distinct properties in sensitivity to perturbation factors and data complexity, hinting at the potential benefits of their fusion. Our research underscores the substantial potential of speaker augmentation, highlighting the importance of in-depth exploration and analysis.

cs.SD↗