Source author record

Sheng Sun

Sheng Sun appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Artificial Intelligence Machine Learning physics.app-ph physics.optics Distributed, Parallel, and Cluster Computing physics.ins-det

Catalog footprint

What is connected

6works

6topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

Rationale-Grounded In-Context Learning for Time Series Reasoning with Multimodal Large Language Models

The underperformance of existing multimodal large language models for time series reasoning lies in the absence of rationale priors that connect temporal observations to their downstream outcomes, which leads models to rely on superficial pattern matching rather than principled reasoning. We therefore propose the rationale-grounded in-context learning for time series reasoning, where rationales work as guiding reasoning units rather than post-hoc explanations, and develop the RationaleTS method. Specifically, we firstly induce label-conditioned rationales, composed of reasoning paths from observable evidence to the potential outcomes. Then, we design the hybrid retrieval by balancing temporal patterns and semantic contexts to retrieve correlated rationale priors for the final in-context inference on new samples. We conduct extensive experiments to demonstrate the effectiveness and efficiency of our proposed RationaleTS on three-domain time series reasoning tasks. We will release our code for reproduction.

preprint2024arXiv

Logits Poisoning Attack in Federated Distillation

Federated Distillation (FD) is a novel and promising distributed machine learning paradigm, where knowledge distillation is leveraged to facilitate a more efficient and flexible cross-device knowledge transfer in federated learning. By optimizing local models with knowledge distillation, FD circumvents the necessity of uploading large-scale model parameters to the central server, simultaneously preserving the raw data on local clients. Despite the growing popularity of FD, there is a noticeable gap in previous works concerning the exploration of poisoning attacks within this framework. This can lead to a scant understanding of the vulnerabilities to potential adversarial actions. To this end, we introduce FDLA, a poisoning attack method tailored for FD. FDLA manipulates logit communications in FD, aiming to significantly degrade model performance on clients through misleading the discrimination of private samples. Through extensive simulation experiments across a variety of datasets, attack scenarios, and FD configurations, we demonstrate that LPA effectively compromises client model accuracy, outperforming established baseline algorithms in this regard. Our findings underscore the critical need for robust defense mechanisms in FD settings to mitigate such adversarial threats.

preprint2022arXiv

Approaching the Fundamental Limit of Orbital Angular Momentum Multiplexing Through a Hologram Metasurface

Establishing and approaching the fundamental limit of orbital angular momentum (OAM) multiplexing are necessary and increasingly urgent for current multiple-input multiple-output research. In this work, we elaborate the fundamental limit in terms of independent scattering channels (or degrees of freedom of scattered fields) through angular-spectral analysis, in conjunction with a rigorous Green function method. The scattering channel limit is universal for arbitrary spatial mode multiplexing, which is launched by a planar electromagnetic device, such as antenna, metasurface, etc, with a predefined physical size. As a proof of concept, we demonstrate both theoretically and experimentally the limit by a metasurface hologram that transforms orthogonal OAM modes to plane-wave modes scattered at critically separated angular-spectral regions. Particularly, a minimax optimization algorithm is applied to suppress angular spectrum aliasing, achieving good performances in both full-wave simulation and experimental measurement at microwave frequencies. This work offers a theoretical upper bound and corresponding approach route for engineering designs of OAM multiplexing.

preprint2022arXiv

Towards Federated Learning against Noisy Labels via Local Self-Regularization

Federated learning (FL) aims to learn joint knowledge from a large scale of decentralized devices with labeled data in a privacy-preserving manner. However, since high-quality labeled data require expensive human intelligence and efforts, data with incorrect labels (called noisy labels) are ubiquitous in reality, which inevitably cause performance degradation. Although a lot of methods are proposed to directly deal with noisy labels, these methods either require excessive computation overhead or violate the privacy protection principle of FL. To this end, we focus on this issue in FL with the purpose of alleviating performance degradation yielded by noisy labels meanwhile guaranteeing data privacy. Specifically, we propose a Local Self-Regularization method, which effectively regularizes the local training process via implicitly hindering the model from memorizing noisy labels and explicitly narrowing the model output discrepancy between original and augmented instances using self distillation. Experimental results demonstrate that our proposed method can achieve notable resistance against noisy labels in various noise levels on three benchmark datasets. In addition, we integrate our method with existing state-of-the-arts and achieve superior performance on the real-world dataset Clothing1M. The code is available at https://github.com/Sprinter1999/FedLSR.

preprint2021arXiv

Manipulation of Orbital Angular Momentum Spectrum Using Shape-Tailored Metasurfaces

Vortex beams carrying orbital angular momentum (OAM) have been widely applied in various electromagnetic, optical, and quantum systems. A tailored OAM spectrum composed of several specific modes as expected holds a promise for expanding the degrees of freedom of the systems. However, such a broadband high-purity tailored spectrum is difficult to be achieved by the present devices, where the broadband amplitude manipulation has not been explored yet. In this work, inspired by the envelope-modulation theory, an elegant and universal way to manipulate the OAM spectrum in wide bandwidth is proposed by using a shape-tailored metasurface. Firstly, the rotating meta-atoms on a triangular lattice are proved to have smaller coupling distortion than that on a square lattice, and this behavior is critical for high-purity vortex spectrum generation by the Pancharatnam-Berry-based metasurfaces. Secondly, a universal modulation relation is established between the spatial arrangement of metasurfaces and the generated vortex beams. Finally, the broadband modulated OAM spectra and the comb-like OAM spectra are theoretically and experimentally demonstrated by the shape-tailored metasurfaces. The proposed amplitude-modulation scheme offers a novel concept and engineering route to manipulate the OAM spectrum in wide bandwidth, which could promote the development of OAM-based applications.

preprint2014arXiv

Design of Wideband Microstrip Filters with Non-Equiripple Responses and Low Sensitivity

This paper presents a novel design procedure for wideband microstrip bandpass filters with non-equiripple filtering frequency responses and low sensitivity. Different from the traditional Chebyshev transfer function filters, the return loss zeros of the proposed non-equiripple filters can be redistributed within the operating passband. For the industrial applications, the proposed filters have a reduced sensitivity to manufacturing errors and exhibit good tolerance control for both specified bandwidth and maximum in-band reflection loss. By deriving the transfer functions, a synthesis approach with a set of non-linear equations can be established according to the specifications such as the bandwidth and predetermined reflection lobes. Without performing any post optimization in the full-wave simulation, the non-equiripple synthesized results have less sensitivity and fractional bandwidth (delta) error in comparison with those obtained from traditional Chebyshev transfer functions with equiripple frequency responses. As design examples, a four-pole bandpass filter with delta=60% and a five-pole bandpass filter with delta=82.5% are designed and fabricated. Measured results show a good agreement with those obtained from the prediction, without any tuning or adjustments.