arXiv subjects

Xiang Sun

Publications and source records attributed to Xiang Sun.

At least 19 recordsLinked to original sources

arXiv2026

Multi-Agent Reinforcement Learning for UAV-Based Chemical Plume Source Localization

Undocumented orphaned wells pose significant health and environmental risks to nearby communities by releasing toxic gases and contaminating water sources, with methane emissions being a primary concern. Traditional survey methods such as magnetometry often fail to detect older wells effectively. In contrast, aerial in-situ sensing using unmanned aerial vehicles (UAVs) offers a promising alternative for methane emission detection and source localization. This study presents a robust and efficient framework based on a multi-agent deep reinforcement learning (MARL) algorithm for the chemical plume source localization (CPSL) problem. The proposed approach leverages virtual anchor nodes to coordinate UAV navigation, enabling collaborative sensing of gas concentrations and wind velocities through onboard and shared measurements. Source identification is achieved by analyzing the historical trajectory of anchor node placements within the plume. Comparative evaluations against the fluxotaxis method demonstrate that the MARL framework achieves superior performance in both localization accuracy and operational efficiency.

eess.SY↗

arXiv2026

Compute Only Once: UG-Separation for Efficient Large Recommendation Models

Driven by scaling laws, recommender systems increasingly rely on larger-scale models to capture complex feature interactions and user behaviors, but this trend also leads to prohibitive training and inference costs. While long-sequence models can reuse user-side computation through KV Caching, such reuse is difficult in TokenMixer-based dense feature interaction architectures, where user and group features are deeply entangled and mixed-up across layers. In this work, we present User-Group Separation (UG-Sep), an industrial large-scale framework that enables user-side computation reusable in TokenMixer-based dense interaction models for the first time. UG-Sep explicitly disentangles user-side and item-side information flows within token-mixing layers, ensuring that a subset of tokens preserves purely user-side representations across layers. This design allows the corresponding per-token computations to be reused across multiple samples, significantly reducing redundant inference cost. To compensate for the potential expressive capacity loss induced by masking, we further propose an Information Compensation strategy that adaptively reconstructs suppressed user-item interactions. Moreover, as UG-Sep substantially reduces user-side FLOPs and exposes memory-bound components, we incorporate W8A16 (8-bit weight, 16-bit activation) weight-only quantization to alleviate memory bandwidth bottlenecks and achieve additional acceleration. We conduct extensive offline evaluations and large-scale online A/B experiments at ByteDance to validate the effectiveness of UG-Sep. Results show that UG-Sep reduces inference latency by up to 20% without causing adverse changes to online user experience and commercial metrics on multiple influential business scenarios compared to TokenMixer at ByteDance, including Douyin Feed Recommendation, Hongguo Feed Recommendation, Chuanshanjia Ads, and Qianchuan Ads.

cs.IR↗

arXiv2025

FusionMAE: large-scale pretrained model to optimize and simplify diagnostic and control of fusion plasma

In magnetically confined fusion device, the complex, multiscale, and nonlinear dynamics of plasmas necessitate the integration of extensive diagnostic systems to effectively monitor and control plasma behaviour. The complexity and uncertainty arising from these extensive systems and their tangled interrelations has long posed a significant obstacle to the acceleration of fusion energy development. In this work, a large-scale model, fusion masked auto-encoder (FusionMAE) is pre-trained to compress the information from 88 diagnostic signals into a concrete embedding, to provide a unified interface between diagnostic systems and control actuators. Two mechanisms are proposed to ensure a meaningful embedding: compression-reduction and missing-signal reconstruction. Upon completion of pre-training, the model acquires the capability for 'virtual backup diagnosis', enabling the inference of missing diagnostic data with 96.7% reliability. Furthermore, the model demonstrates three emergent capabilities: automatic data analysis, universal control-diagnosis interface, and enhancement of control performance on multiple tasks. This work pioneers large-scale AI model integration in fusion energy, demonstrating how pre-trained embeddings can simplify the system interface, reducing necessary diagnostic systems and optimize operation performance for future fusion reactors.

physics.plasm-ph↗

arXiv2025

LONGER: Scaling Up Long Sequence Modeling in Industrial Recommenders

Modeling ultra-long user behavior sequences is critical for capturing both long- and short-term preferences in industrial recommender systems. Existing solutions typically rely on two-stage retrieval or indirect modeling paradigms, incuring upstream-downstream inconsistency and computational inefficiency. In this paper, we present LONGER, a Long-sequence Optimized traNsformer for GPU-Efficient Recommenders. LONGER incorporates (i) a global token mechanism for stabilizing attention over long contexts, (ii) a token merge module with lightweight InnerTransformers and hybrid attention strategy to reduce quadratic complexity, and (iii) a series of engineering optimizations, including training with mixed-precision and activation recomputation, KV cache serving, and the fully synchronous model training and serving framework for unified GPU-based dense and sparse parameter updates. LONGER consistently outperforms strong baselines in both offline metrics and online A/B testing in both advertising and e-commerce services at ByteDance, validating its consistent effectiveness and industrial-level scaling laws. Currently, LONGER has been fully deployed at more than 10 influential scenarios at ByteDance, serving billion users.

cs.IR↗

arXiv2025

Sky of Unlearning (SoUL): Rewiring Federated Machine Unlearning via Selective Pruning

The Internet of Drones (IoD), where drones collaborate in data collection and analysis, has become essential for applications such as surveillance and environmental monitoring. Federated learning (FL) enables drones to train machine learning models in a decentralized manner while preserving data privacy. However, FL in IoD networks is susceptible to attacks like data poisoning and model inversion. Federated unlearning (FU) mitigates these risks by eliminating adversarial data contributions, preventing their influence on the model. This paper proposes sky of unlearning (SoUL), a federated unlearning framework that efficiently removes the influence of unlearned data while maintaining model performance. A selective pruning algorithm is designed to identify and remove neurons influential in unlearning but minimally impact the overall performance of the model. Simulations demonstrate that SoUL outperforms existing unlearning methods, achieves accuracy comparable to full retraining, and reduces computation and communication overhead, making it a scalable and efficient solution for resource-constrained IoD networks.

cs.LG↗

arXiv2024

Dynamic Client Clustering, Bandwidth Allocation, and Workload Optimization for Semi-synchronous Federated Learning

Federated Learning (FL) revolutionizes collaborative machine learning among Internet of Things (IoT) devices by enabling them to train models collectively while preserving data privacy. FL algorithms fall into two primary categories: synchronous and asynchronous. While synchronous FL efficiently handles straggler devices, it can compromise convergence speed and model accuracy. In contrast, asynchronous FL allows all devices to participate but incurs high communication overhead and potential model staleness. To overcome these limitations, the semi-synchronous FL framework introduces client tiering based on computing and communication latencies. Clients in different tiers upload their local models at distinct frequencies, striking a balance between straggler mitigation and communication costs. Enter the DecantFed algorithm (Dynamic client clustering, bandwidth allocation, and local training for semi-synchronous Federated learning), a dynamic solution that optimizes client clustering, bandwidth allocation, and local training workloads to maximize data sample processing rates. Additionally, DecantFed adapts client learning rates according to their tiers, addressing the model staleness problem. The algorithm's performance shines in extensive simulations using benchmark datasets, including MNIST and CIFAR-10, under independent and identically distributed (IID) and non-IID scenarios. DecantFed outpaces FedAvg and FedProx in terms of convergence speed and delivers a remarkable minimum 28% boost in model accuracy compared to FedProx.

cs.DC↗

arXiv2023

Octree-based hierarchical sampling optimization for the volumetric super-resolution of scientific data

When introducing physics-constrained deep learning solutions to the volumetric super-resolution of scientific data, the training is challenging to converge and always time-consuming. We propose a new hierarchical sampling method based on octree to solve these difficulties. In our approach, scientific data is preprocessed before training, and a hierarchical octree-based data structure is built to guide sampling on the latent context grid. Each leaf node in the octree corresponds to an indivisible subblock of the volumetric data. The dimensions of the subblocks are different, making the number of sample points in each randomly cropped training data block to be adaptive. We reconstruct the octree at intervals according to loss distribution to perform the multi-stage training. With the Rayleigh-B\'enard convection problem, we deploy our method to state-of-the-art models. We constructed adequate experiments to evaluate the training performance and model accuracy of our method. Experiments indicate that our sampling optimization improves the convergence performance of physics-constrained deep learning super-resolution solutions. Furthermore, the sample points and training time are significantly reduced with no drop in model accuracy. We also test our method in training tasks of other deep neural networks, and the results show our sampling optimization has extensive effectiveness and applicability. The code is publicly available at https://github.com/xinjiewang/octree-based_sampling.

physics.comp-ph↗

arXiv2023

GraphGANFed: A Federated Generative Framework for Graph-Structured Molecules Towards Efficient Drug Discovery

Recent advances in deep learning have accelerated its use in various applications, such as cellular image analysis and molecular discovery. In molecular discovery, a generative adversarial network (GAN), which comprises a discriminator to distinguish generated molecules from existing molecules and a generator to generate new molecules, is one of the premier technologies due to its ability to learn from a large molecular data set efficiently and generate novel molecules that preserve similar properties. However, different pharmaceutical companies may be unwilling or unable to share their local data sets due to the geo-distributed and sensitive nature of molecular data sets, making it impossible to train GANs in a centralized manner. In this paper, we propose a Graph convolutional network in Generative Adversarial Networks via Federated learning (GraphGANFed) framework, which integrates graph convolutional neural Network (GCN), GAN, and federated learning (FL) as a whole system to generate novel molecules without sharing local data sets. In GraphGANFed, the discriminator is implemented as a GCN to better capture features from molecules represented as molecular graphs, and FL is used to train both the discriminator and generator in a distributive manner to preserve data privacy. Extensive simulations are conducted based on the three bench-mark data sets to demonstrate the feasibility and effectiveness of GraphGANFed. The molecules generated by GraphGANFed can achieve high novelty (=100) and diversity (> 0.9). The simulation results also indicate that 1) a lower complexity discriminator model can better avoid mode collapse for a smaller data set, 2) there is a tradeoff among different evaluation metrics, and 3) having the right dropout ratio of the generator and discriminator can avoid mode collapse.

cs.LG↗

arXiv2023

Effort Discrimination and Curvature of Contest Technology in Conflict Networks

In a model of interconnected conflicts on a network, we compare the equilibrium effort profiles and payoffs under two scenarios: uniform effort (UE) in which each contestant is restricted to exert the same effort across all the battles she participates, and discriminatory effort (DE) in which such a restriction is lifted. When the contest technology in each battle is of Tullock form, a surprising neutrality result holds within the class of semi-symmetric conflict network structures: both the aggregate actions and equilibrium payoffs under two regimes are the same. We also show that, in some sense, the Tullock form is necessary for such a neutrality result. Moving beyond the Tullock family, we further demonstrate how the curvature of contest technology shapes the welfare and effort effects.

econ.TH↗

arXiv2022

Latency Aware Semi-synchronous Client Selection and Model Aggregation for Wireless Federated Learning

Federated learning (FL) is a collaborative machine learning framework that requires different clients (e.g., Internet of Things devices) to participate in the machine learning model training process by training and uploading their local models to an FL server in each global iteration. Upon receiving the local models from all the clients, the FL server generates a global model by aggregating the received local models. This traditional FL process may suffer from the straggler problem in heterogeneous client settings, where the FL server has to wait for slow clients to upload their local models in each global iteration, thus increasing the overall training time. One of the solutions is to set up a deadline and only the clients that can upload their local models before the deadline would be selected in the FL process. This solution may lead to a slow convergence rate and global model overfitting issues due to the limited client selection. In this paper, we propose the Latency awarE Semi-synchronous client Selection and mOdel aggregation for federated learNing (LESSON) method that allows all the clients to participate in the whole FL process but with different frequencies. That is, faster clients would be scheduled to upload their models more frequently than slow clients, thus resolving the straggler problem and accelerating the convergence speed, while avoiding model overfitting. Also, LESSON is capable of adjusting the tradeoff between the model accuracy and convergence rate by varying the deadline. Extensive simulations have been conducted to compare the performance of LESSON with the other two baseline methods, i.e., FedAvg and FedCS. The simulation results demonstrate that LESSON achieves faster convergence speed than FedAvg and FedCS, and higher model accuracy than FedCS.

cs.DC↗

arXiv2022

Deep Reinforcement Learning for Online Latency Aware Workload Offloading in Mobile Edge Computing

Owing to the resource-constrained feature of Internet of Things (IoT) devices, offloading tasks from IoT devices to the nearby mobile edge computing (MEC) servers can not only save the energy of IoT devices but also reduce the response time of executing the tasks. However, offloading a task to the nearest MEC server may not be the optimal solution due to the limited computing resources of the MEC server. Thus, jointly optimizing the offloading decision and resource management is critical, but yet to be explored. Here, offloading decision refers to where to offload a task and resource management implies how much computing resource in an MEC server is allocated to a task. By considering the waiting time of a task in the communication and computing queues (which are ignored by most of the existing works) as well as tasks priorities, we propose the \ul{D}eep reinforcement l\ul{E}arning based offloading de\ul{C}ision and r\ul{E}source manageme\ul{NT} (DECENT) algorithm, which leverages the advantage actor critic method to optimize the offloading decision and computing resource allocation for each arriving task in real-time such that the cumulative weighted response time can be minimized. The performance of DECENT is demonstrated via different experiments.

cs.DC↗

arXiv2021

Backhaul-aware Drone Base Station Placement and Resource Management for FSO-based Drone-assisted Mobile Networks

In drone-assisted mobile networks, Drone-mounted Base Stations (DBSs) are responsively and flexibly deployed over any Places of Interest (PoI), such as sporadic hotspots and disaster-struck areas, where the existing mobile network infrastructure is unable to provide wireless coverage. In this paper, a DBS is an aerial base station to relay traffic between a nearby Macro Base Station (MBS) and the users. In addition, Free Space Optics (FSO) is applied as the backhauling solution to significantly increase the capacity of the backhaul link between an MBS and a DBS. Most of the existing DBS placement solutions assume the FSO-based backhaul link provides sufficient link capacity, which may not be true, especially when a DBS is placed far away from an MBS (e.g., > 10 km in disaster-struck areas) or in a bad weather condition. In this paper, we formulate a problem to jointly optimize bandwidth allocation and DBS placement by considering the FSO-based backhaul link capacity constraint. A Backhaul awaRe bandwidth allOcAtion and DBS placement (BROAD) algorithm is designed to efficiently solve the problem, and the performance of the algorithm is demonstrated via extensive simulations.

cs.NI↗

arXiv2021

Nature of $X(2370)$

We address the nature of the $X(2370)$ resonance observed in the $J/\psi$ radiative decays, $J/\psi\rightarrow\gamma K^{+} K^{-}\eta'$, $J/\psi\rightarrow\gamma K_S K_S\eta'$ and $J/\psi\rightarrow\gamma \pi^{+}\pi^{-}\eta'$. By studying the invariant mass spectra we confirm that decays of the $X(2370)$ into three pseudo-scalars are well described by an effective chiral Lagrangian. We extract the branching ratio of $J/\psi\to X(2370)\gamma$ and show that it is an order of magnitude larger compared to the glueball production rate predicted by lattice QCD. This indicates that $X(2370)$ is not likely to be a glueball candidate.

hep-ph↗

arXiv2021

Pole analysis on the doubly charmed meson in $D^0D^0\pi^+$ mass spectrum

In this paper we study the scattering amplitudes of $D^{0}D^{0}\pi^+$-$D^{*+}D^{0}$ coupled channels based on $K$-matrix within the Chew-Mandelstam formalism. The $D^{0}D^{0}\pi^{+}$ invariant mass spectrum of LHCb is fitted and the pole parameters of the $T_{cc}^+$ are extracted. The analysis of pole behavior suggests that the $T_{cc}^+$ may originate from a $D^{*+}D^{0}$ virtual state and is formed as a result of an interplay between an attractive interaction between $D^0$ and $D^{*+}$ and coupling to $D^{0}D^0\pi^+$ channel.

hep-ph↗

arXiv2021

Characterization of equilibrium existence and purification in general Bayesian games

This paper studies Bayesian games with general action spaces, correlated types and interdependent payoffs. We introduce the condition of ``decomposable coarser payoff-relevant information'', and show that this condition is both sufficient and necessary for the existence of pure-strategy equilibria and purification from behavioral strategies. As a consequence of our purification method, a new existence result on pure-strategy equilibria is also obtained for discontinuous Bayesian games. Illustrative applications of our results to oligopolistic competitions and all-pay auctions are provided.

econ.GN↗

arXiv2021

Directly profiling the dark-state transition density via scanning tunneling microscope

The molecular dark state participates in many important photon-induced processes, yet is typically beyond the optical-spectroscopic measurement due to the forbidden transition dictated by the selection rule. In this work, we propose to use the scanning tunneling microscope (STM) as an incisive tool to directly profile the dark-state transition density of a single molecule, taking advantage of the localized static electronic field near the metal tip. The detection of dark state is achieved by measuring the fluorescence from a higher bright state to the ground state with assistant optical pumping. The current proposal shall bring new methodology to study the single-molecule properties in the electro-optical devices and the light-assisted biological processes.

quant-ph↗

arXiv2021

CTRAMER: An open source software package for correlating interfacial charge transfer rate constants with donor acceptor geometries in organic photovoltaic materials

In this paper we present CTRAMER (Charge Transfer RAtes from Molecular dynamics, Electronic structure, and Rate theory), an open source software package for calculating interfacial charge transfer (CT) rate constants in organic photovoltaic (OPV) materials based on ab initio calculations and molecular dynamics simulations. The software is based on identifying representative donor acceptor geometries within interfacial structures obtained from molecular dynamics simulation of donor acceptor blends and calculating the corresponding Fermi s golden rule CT rate constants within the framework of the linearized semiclassical approximation. While the methods used are well established, the integration of these state of the art ideas from different disciplines to study photoinduced CT between excited states and explicit environment, in our opinion, makes this package unique and innovative. The software also provides tools for plotting other observables of interest. After outlining the features and implementation details, usage and performance of the software are demonstrated with results from an example OPV system.

cond-mat.mtrl-sci↗

arXiv2021

An HPC-Based Hydrothermal Finite Element Simulator for Modeling Underground Response to Community-Scale Geothermal Energy Production

Geothermal heat, as renewable energy, shows great advantage with respect to its environmental impact due to its significantly lower CO2 emissions than conventional fossil fuel. Open and closed-loop geothermal heat pumps, which utilize shallow geothermal systems, are an efficient technology for cooling and heating buildings, especially in urban areas. Integrated use of geothermal energy technologies for district heating, cooling, and thermal energy storage can be applied to optimize the subsurface for communities to provide them with multiple sustainable energy and community resilience benefits. The utilization of the subsurface resources may lead to a variation in the underground environment, which might further impact local environmental conditions. However, very few simulators can handle such a highly complex set of coupled computations on a regional or city scale. We have developed high-performance computing (HPC) based hydrothermal finite element (FE) simulator that can simulate the subsurface and its hydrothermal conditions at a scale of tens of km. The HPC simulator enables us to investigate the subsurface thermal and hydrologic response to the built underground environment (such as basements and subways) at the community scale. In this study, a coupled hydrothermal simulator is developed based on the open-source finite element library deal.II. The HPC simulator was validated by comparing the results of a benchmark case study against COMSOL Multiphysics, in which Aquifer Thermal Energy Storage (ATES) is modeled and a process of heat injection into ATES is simulated. The use of an energy pile system at the Treasure Island redevelopment site (San Francisco, CA, USA) was selected as a case study to demonstrate the HPC capability of the developed simulator. The simulator is capable of modeling multiple city-scale geothermal scenarios in a reasonable amount of time.

cs.DC↗