arXiv subjects

Sheng Luo

Publications and source records attributed to Sheng Luo.

At least 19 recordsLinked to original sources

arXiv2026

MISO: Model-Internal-State-Guided Optimization for Ranking Models

Ranking models are repeatedly refined within established model families, yet the choice of which component to scale, replace, or retire is often guided by expensive trial-and-error. We present Model Internal State Optimization (MISO), a systems workflow that uses model internal states (MIS), including parameters, activations, gradients, and normalization statistics, to prioritize such local optimization decisions. MISO extracts MIS from a trained ranking model, aggregates them into ranking, alignment, and comparison signals, and converts those signals into a small set of interpretable candidate edits. Because MIS are re-extracted after each retraining cycle, MISO naturally supports an adaptive optimization workflow that tracks evolving model behavior as data distributions and system requirements shift over time. In an ads ranking case study, MISO improves normalized entropy while requiring substantially fewer validation runs than expert-driven and black-box scaling workflows, offering a practical middle ground between manual tuning and opaque automated search.

cs.IR↗

arXiv2025

CPMamba: Selective State Space Models for MIMO Channel Prediction in High-Mobility Environments

Channel prediction is a key technology for improving the performance of various functions such as precoding, adaptive modulation, and resource allocation in MIMO-OFDM systems. Especially in high-mobility scenarios with fast time-varying channels, it is crucial for resisting channel aging and ensuring communication quality. However, existing methods suffer from high complexity and the inability to accurately model the temporal variations of channels. To address this issue, this paper proposes CPMamba -- an efficient channel prediction framework based on the selective state space model. The proposed CPMamba architecture extracts features from historical channel state information (CSI) using a specifically designed feature extraction and embedding network and employs stacked residual Mamba modules for temporal modeling. By leveraging an input-dependent selective mechanism to dynamically adjust state transitions, it can effectively capture the long-range dependencies between the CSIs while maintaining a linear computational complexity. Simulation results under the 3GPP standard channel model demonstrate that CPMamba achieves state-of-the-art prediction accuracy across all scenarios, along with superior generalization and robustness. Compared to existing baseline models, CPMamba reduces the number of parameters by approximately 50 percent while achieving comparable or better performance, thereby significantly lowering the barrier for practical deployment.

eess.SP↗

arXiv2025

Demographic Distribution Matching between real world and virtual phantom population

Virtual imaging trials (VITs) offer scalable and cost-effective tools for evaluating imaging systems and protocols. However, their translational impact depends on rigorous comparability between virtual and real-world populations. This study introduces DISTINCT (Distributional Subsampling for Covariate-Targeted Alignment), a statistical framework for selecting demographically aligned subsamples from large clinical datasets to support robust comparisons with virtual cohorts. We applied DISTINCT to the National Lung Screening Trial (NLST) and a companion virtual trial dataset (VLST). The algorithm jointly aligned typical continuous (age, BMI) and categorical (sex, race, ethnicity) variables by constructing multidimensional bins based on discretized covariates. For a given target size, DISTINCT samples individuals to match the joint demographic distribution of the reference population. We evaluated the demographic similarity between VLST and progressively larger NLST subsamples using Wasserstein and Kolmogorov-Smirnov (K-S) distances to identify the maximal subsample size with acceptable alignment. The algorithm identified a maximal aligned NLST subsample of 9,974 participants, preserving demographic similarity to the VLST population. Receiver operating characteristic (ROC) analysis using risk scores for lung cancer detection showed that area under the curve (AUC) estimates stabilized beyond 6,000 participants, confirming the sufficiency of aligned subsamples for virtual imaging trial evaluation. Stratified AUC analysis revealed substantial performance variation across demographic subgroups, reinforcing the importance of covariate alignment in comparative studies.

stat.AP↗

arXiv2025

A novel smoothing-based goodness-of-fit test of covariance for multivariate sparse functional data

Accurately specifying covariance structures is critical for valid inference in longitudinal and functional data analysis, particularly when data are sparsely observed. In this study, we develop a global goodness-of-fit test to assess parametric covariance structures in multivariate sparse functional data. Our contribution is twofold. First, we extend the univariate goodness-of-fit test proposed by Chen et al. (2019) to better accommodate sparse data by improving error variance estimation and applying positive semi-definite smoothing to covariance estimation. These corrections ensure appropriate Type I error control under sparse designs. Second, we introduce a multivariate extension of the improved test that jointly evaluates covariance structures across multiple outcomes, employing novel test statistics based on the maximum and $\ell_2$ norms to account for inter-outcome dependencies and enhance statistical power. Through extensive simulation studies, we demonstrate that the proposed methods maintain proper Type I error rates and achieve greater power than univariate tests with multiple testing adjustments. Applications to longitudinal neuroimaging and clinical data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Parkinson's Progression Marker Initiative (PPMI) illustrate the practical utility of the proposed methods for evaluating covariance structures in sparse multivariate longitudinal data.

stat.ME↗

arXiv2025

Ensemble Survival Analysis for Preclinical Cognitive Decline Prediction in Alzheimer's Disease Using Longitudinal Biomarkers

Predicting the risk of clinical progression from cognitively normal (CN) status to mild cognitive impairment (MCI) or Alzheimer's disease (AD) is critical for early intervention in Alzheimer's disease (AD). Traditional survival models often fail to capture complex longitudinal biomarker patterns associated with disease progression. We propose an ensemble survival analysis framework integrating multiple survival models to improve early prediction of clinical progression in initially cognitively normal individuals. We analyzed longitudinal biomarker data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort, including 721 participants, limiting analysis to up to three visits (baseline, 6-month follow-up, 12-month follow-up). Of these, 142 (19.7%) experienced clinical progression to MCI or AD. Our approach combined penalized Cox regression (LASSO, Elastic Net) with advanced survival models (Random Survival Forest, DeepSurv, XGBoost). Model predictions were aggregated using ensemble averaging and Bayesian Model Averaging (BMA). Predictive performance was assessed using Harrell's concordance index (C-index) and time-dependent area under the curve (AUC). The ensemble model achieved a peak C-index of 0.907 and an integrated time-dependent AUC of 0.904, outperforming baseline-only models (C-index 0.608). One follow-up visit after baseline significantly improved prediction accuracy (48.1% C-index, 48.2% AUC gains), while adding a second follow-up provided only marginal gains (2.1% C-index, 2.7% AUC). Our ensemble survival framework effectively integrates diverse survival models and aggregation techniques to enhance early prediction of preclinical AD progression. These findings highlight the importance of leveraging longitudinal biomarker data, particularly one follow-up visit, for accurate risk stratification and personalized intervention strategies.

stat.AP↗

arXiv2025

Universal Incremental Learning: Mitigating Confusion from Inter- and Intra-task Distribution Randomness

Incremental learning (IL) aims to overcome catastrophic forgetting of previous tasks while learning new ones. Existing IL methods make strong assumptions that the incoming task type will either only increases new classes or domains (i.e. Class IL, Domain IL), or increase by a static scale in a class- and domain-agnostic manner (i.e. Versatile IL (VIL)), which greatly limit their applicability in the unpredictable and dynamic wild. In this work, we investigate $\textbf{Universal Incremental Learning (UIL)}$, where a model neither knows which new classes or domains will increase along sequential tasks, nor the scale of the increments within each task. This uncertainty prevents the model from confidently learning knowledge from all task distributions and symmetrically focusing on the diverse knowledge within each task distribution. Consequently, UIL presents a more general and realistic IL scenario, making the model face confusion arising from inter-task and intra-task distribution randomness. To $\textbf{Mi}$tigate both $\textbf{Co}$nfusion, we propose a simple yet effective framework for UIL, named $\textbf{MiCo}$. At the inter-task distribution level, we employ a multi-objective learning scheme to enforce accurate and deterministic predictions, and its effectiveness is further enhanced by a direction recalibration module that reduces conflicting gradients. Moreover, at the intra-task distribution level, we introduce a magnitude recalibration module to alleviate asymmetrical optimization towards imbalanced class distribution. Extensive experiments on three benchmarks demonstrate the effectiveness of our method, outperforming existing state-of-the-art methods in both the UIL scenario and the VIL scenario. Our code will be available at $\href{https://github.com/rolsheng/UIL}{here}$.

cs.CV↗

arXiv2025

Power and sample size calculation for multivariate longitudinal trials using the longitudinal rank sum test

Neurodegenerative diseases such as Alzheimer's and Parkinson's often exhibit complex, multivariate longitudinal outcomes that require advanced statistical methods to comprehensively evaluate treatment efficacy. The Longitudinal Rank Sum Test (LRST) offers a nonparametric framework to assess global treatment effects across multiple longitudinal endpoints without requiring multiplicity corrections. This study develops a robust methodology for power and sample size estimation specific to the LRST, integrating theoretical derivations, asymptotic properties, and practical estimation techniques. Validation through numerical simulations demonstrates the accuracy of the proposed methods, while real-world applications to clinical trials in Alzheimer's and Parkinson's disease highlight their practical significance. This framework facilitates the design of efficient, well-powered trials, advancing the evaluation of treatments for complex diseases with multivariate longitudinal outcomes.

stat.ME↗

arXiv2024

A novel longitudinal rank-sum test for multiple primary endpoints in clinical trials: Applications to neurodegenerative disorders

Neurodegenerative disorders such as Alzheimer's disease (AD) present a significant global health challenge, characterized by cognitive decline, functional impairment, and other debilitating effects. Current AD clinical trials often assess multiple longitudinal primary endpoints to comprehensively evaluate treatment efficacy. Traditional methods, however, may fail to capture global treatment effects, require larger sample sizes due to multiplicity adjustments, and may not fully exploit multivariate longitudinal data. To address these limitations, we introduce the Longitudinal Rank Sum Test (LRST), a novel nonparametric rank-based omnibus test statistic. The LRST enables a comprehensive assessment of treatment efficacy across multiple endpoints and time points without multiplicity adjustments, effectively controlling Type I error while enhancing statistical power. It offers flexibility against various data distributions encountered in AD research and maximizes the utilization of longitudinal data. Extensive simulations and real-data applications demonstrate the LRST's performance, underscoring its potential as a valuable tool in AD clinical trials.

stat.ME↗

arXiv2024

A non-parametric U-statistic testing approach for multi-arm clinical trials with multivariate longitudinal data

Randomized clinical trials (RCTs) often involve multiple longitudinal primary outcomes to comprehensively assess treatment efficacy. The Longitudinal Rank-Sum Test (LRST), a robust U-statistics-based, non-parametric, rank-based method, effectively controls Type I error and enhances statistical power by leveraging the temporal structure of the data without relying on distributional assumptions. However, the LRST is limited to two-arm comparisons. To address the need for comparing multiple doses against a control group in many RCTs, we extend the LRST to a multi-arm setting. This novel multi-arm LRST provides a flexible and powerful approach for evaluating treatment efficacy across multiple arms and outcomes, with a strong capability for detecting the most effective dose in multi-arm trials. Extensive simulations demonstrate that this method maintains excellent Type I error control while providing greater power compared to the two-arm LRST with multiplicity adjustments. Application to the Bapineuzumab (Bapi) 301 trial further validates the multi-arm LRST's practical utility and robustness, confirming its efficacy in complex clinical trial analyses.

stat.ME↗

arXiv2024

Infinite time horizon stochastic recursive control problems with jumps: dynamic programming and stochastic verification theorems

This paper is devoted to studying an infinite time horizon stochastic recursive control problem with jumps, where infinite time horizon stochastic differential equation and backward stochastic differential equation with jumps describe the state process and cost functional, respectively. For this, the first is to explore the wellposedness and regularity of these two equations in $L^p$-sense ($p\geq2$). By establishing the dynamic programming principle, we relate the value function of the control problem with integral-partial differential equation of HJB type in the sense of viscosity solutions. On the other hand, stochastic verification theorems are also studied to provide sufficient conditions to verify the optimality of the given admissible controls. Such a study is carried out in the framework of classical solutions but also in that of viscosity solutions. Our work emphasizes important differences from the approach for finite time horizon problems. In particular, we have to work in an $L^p$-setting for $p>4$ in order to study the verification theorem in viscosity sense.

math.OC↗

arXiv2024

Virtual Lung Screening Trial (VLST): An In Silico Study Inspired by the National Lung Screening Trial for Lung Cancer Detection

Clinical imaging trials play a crucial role in advancing medical innovation but are often costly, inefficient, and ethically constrained. Virtual Imaging Trials (VITs) present a solution by simulating clinical trial components in a controlled, risk-free environment. The Virtual Lung Screening Trial (VLST), an in silico study inspired by the National Lung Screening Trial (NLST), illustrates the potential of VITs to expedite clinical trials, minimize risks to participants, and promote optimal use of imaging technologies in healthcare. This study aimed to show that a virtual imaging trial platform could investigate some key elements of a major clinical trial, specifically the NLST, which compared Computed tomography (CT) and chest radiography (CXR) for lung cancer screening. With simulated cancerous lung nodules, a virtual patient cohort of 294 subjects was created using XCAT human models. Each virtual patient underwent both CT and CXR imaging, with deep learning models, the AI CT-Reader and AI CXR-Reader, acting as virtual readers to perform recall patients with suspicion of lung cancer. The primary outcome was the difference in diagnostic performance between CT and CXR, measured by the Area Under the Curve (AUC). The AI CT-Reader showed superior diagnostic accuracy, achieving an AUC of 0.92 (95% CI: 0.90-0.95) compared to the AI CXR-Reader's AUC of 0.72 (95% CI: 0.67-0.77). Furthermore, at the same 94% CT sensitivity reported by the NLST, the VLST specificity of 73% was similar to the NLST specificity of 73.4%. This CT performance highlights the potential of VITs to replicate certain aspects of clinical trials effectively, paving the way toward a safe and efficient method for advancing imaging-based diagnostics.

eess.IV↗

arXiv2024

Delving into Multi-modal Multi-task Foundation Models for Road Scene Understanding: From Learning Paradigm Perspectives

Foundation models have indeed made a profound impact on various fields, emerging as pivotal components that significantly shape the capabilities of intelligent systems. In the context of intelligent vehicles, leveraging the power of foundation models has proven to be transformative, offering notable advancements in visual understanding. Equipped with multi-modal and multi-task learning capabilities, multi-modal multi-task visual understanding foundation models (MM-VUFMs) effectively process and fuse data from diverse modalities and simultaneously handle various driving-related tasks with powerful adaptability, contributing to a more holistic understanding of the surrounding scene. In this survey, we present a systematic analysis of MM-VUFMs specifically designed for road scenes. Our objective is not only to provide a comprehensive overview of common practices, referring to task-specific models, unified multi-modal models, unified multi-task models, and foundation model prompting techniques, but also to highlight their advanced capabilities in diverse learning paradigms. These paradigms include open-world understanding, efficient transfer for road scenes, continual learning, interactive and generative capability. Moreover, we provide insights into key challenges and future trends, such as closed-loop driving systems, interpretability, embodied driving agents, and world models. To facilitate researchers in staying abreast of the latest developments in MM-VUFMs for road scenes, we have established a continuously updated repository at https://github.com/rolsheng/MM-VUFM4DS

cs.CV↗

arXiv2023

On Designing Multi-UAV aided Wireless Powered Dynamic Communication via Hierarchical Deep Reinforcement Learning

This paper proposes a novel design on the wireless powered communication network (WPCN) in dynamic environments under the assistance of multiple unmanned aerial vehicles (UAVs). Unlike the existing studies, where the low-power wireless nodes (WNs) often conform to the coherent harvest-then-transmit protocol, under our newly proposed double-threshold based WN type updating rule, each WN can dynamically and repeatedly update its WN type as an E-node for non-linear energy harvesting over time slots or an I-node for transmitting data over sub-slots. To maximize the total transmission data size of all the WNs over T slots, each of the UAVs individually determines its trajectory and binary wireless energy transmission (WET) decisions over times slots and its binary wireless data collection (WDC) decisions over sub-slots, under the constraints of each UAV's limited on-board energy and each WN's node type updating rule. However, due to the UAVs' tightly-coupled trajectories with their WET and WDC decisions, as well as each WN's time-varying battery energy, this problem is difficult to solve optimally. We then propose a new multi-agent based hierarchical deep reinforcement learning (MAHDRL) framework with two tiers to solve the problem efficiently, where the soft actor critic (SAC) policy is designed in tier-1 to determine each UAV's continuous trajectory and binary WET decision over time slots, and the deep-Q learning (DQN) policy is designed in tier-2 to determine each UAV's binary WDC decisions over sub-slots under the given UAV trajectory from tier-1. Both of the SAC policy and the DQN policy are executed distributively at each UAV. Finally, extensive simulation results are provided to validate the outweighed performance of the proposed MAHDRL approach over various state-of-the-art benchmarks.

cs.NI↗

arXiv2023

Ground-State Probabilistic Logic with the Simplest Binary Energy Landscape for Probabilistic Computing

We investigate the ground-state probabilistic logic based on a binary energy landscape (GSPL-BEL) model, implementing the many-body interactions within Ising model cells. The GSPL-BEL model offers a simplified binary energy landscape, enabling the conversion of traditional CMOS-based logic into a probabilistic graphical representation based on desired truth tables. Stochastic Ising cells, coupled with generic probabilistic devices exhibiting sigmoidal electrical responses, serve as the building blocks of the GSPL-BEL. Multi-body interactions are realized through cascaded CMOS-based XNOR gates and a passive resistor network. Through circuit simulations of three-node, four-node, and five-node systems, the functionality of the GSPL-BEL model is verified in forward, reverse, and partial-reverse operating modes, and applied to various arithmetic tasks. The many-body effect provides additional degrees of freedom in describing the system's energy function, resulting in distinct energy levels for valid and invalid states. This observation is supported by the binarized probability distribution observed in the steady state of the probabilistic circuits. Furthermore, compared to conventional combinatorial logic circuits, the GSPL-BEL-based circuit design requires a minimal number of probabilistic devices, as demonstrated in the invertible multiplier/integer factorizer circuit. These findings highlight the potential of the GSPL-BEL model for future high-performance logic circuit designs leveraging probabilistic devices.

physics.app-ph↗

arXiv2023

Power and sample size calculation of two-sample projection-based testing for sparsely observed functional data

Projection-based testing for mean trajectory differences in two groups of irregularly and sparsely observed functional data has garnered significant attention in the literature because it accommodates a wide spectrum of group differences and (non-stationary) covariance structures. This article presents the derivation of the theoretical power function and the introduction of a comprehensive power and sample size (PASS) calculation toolkit tailored to the projection-based testing method developed by Wang (2021). Our approach accommodates a wide spectrum of group difference scenarios and a broad class of covariance structures governing the underlying processes. Through extensive numerical simulation, we demonstrate the robustness of this testing method by showcasing that its statistical power remains nearly unaffected even when a certain percentage of observations are missing, rendering it 'missing-immune'. Furthermore, we illustrate the practical utility of this test through analysis of two randomized controlled trials of Parkinson's disease. To facilitate implementation, we provide a user-friendly R package fPASS, complete with a detailed vignette to guide users through its practical application. We anticipate that this article will significantly enhance the usability of this potent statistical tool across a range of biostatistical applications, with a particular focus on its relevance in the design of clinical trials.

stat.ME↗

arXiv2023

Real Space Characterization of Nonlinear Hall Effect in Confined Directions

The nonlinear Hall effect (NLHE) is a phenomenon which could produce a transverse Hall voltage in a time-reversal-invariant material. Here, we report the real space characterization of NLHE evaluated through quantum transport in TaIrTe4 nanoribbon without the explicit Berry curvature dipole (BCD) information. We first characterize the NLHE in both transverse confined directions in global-level measurement. The impact of quantum confinement in NLHE is evaluated by adjusting the width of nanoribbons. Then, the probing area is trimmed to the atomic scale to evaluate the local texture, where we discover its unique patterns among the probed atomic groups for the first time. The analysis of charge distribution reveals the connections between NLHE's local patterns and its non-centrosymmetric nature, rendering nearly an order of Hall voltage enhancement through probe positioning. Our work paves the way to expand the range of NLHE study and unveil its physics in more versatile material systems.

cond-mat.mes-hall↗

arXiv2023

Multi-Agent Graph Reinforcement Learning based On-Demand Wireless Energy Transfer in Multi-UAV-aided IoT Network

This paper proposes a new on-demand wireless energy transfer (WET) scheme of multiple unmanned aerial vehicles (UAVs). Unlike the existing studies that simply pursuing the total or the minimum harvested energy maximization at the Internet of Things (IoT) devices, where the IoT devices' own energy requirements are barely considered, we propose a new metric called the hungry-level of energy (HoE), which reflects the time-varying energy demand of each IoT device based on the energy gap between its required energy and the harvested energy from the UAVs. With the purpose to minimize the overall HoE of the IoT devices whose energy requirements are not satisfied, we optimally determine all the UAVs' trajectories and WET decisions over time, under the practical mobility and energy constraints of the UAVs. Although the proposed problem is of high complexity to solve, by excavating the UAVs' self-attentions for their collaborative WET, we propose the multiagent graph reinforcement learning (MAGRL) based approach. Through the offline training of the MAGRL model, where the global training at the central controller guides the local training at each UAV agent, each UAV then distributively determines its trajectory and WET based on the well-trained local neural networks. Simulation results show that the proposed MAGRL-based approach outperforms various benchmarks for meeting the IoT devices' energy requirements.

cs.IT↗

arXiv2023

Stochastic representation for solutions of a system of coupled HJB-Isaacs equations with integral-partial operators

In this paper, we focus on the stochastic representation of a system of coupled Hamilton-Jacobi-Bellman-Isaacs (HJB-Isaacs (HJBI), for short) equations which is in fact a system of coupled Isaacs' type integral-partial differential equation. For this, we introduce an associated zero-sum stochastic differential game, where the state process is described by a classical stochastic differential equation (SDE, for short) with jumps, and the cost functional of recursive type is defined by a new type of backward stochastic differential equation (BSDE, for short) with two Poisson random measures, whose wellposedness and a prior estimate as well as the comparison theorem are investigated for the first time. One of the Poisson random measures appearing in the SDE and the BSDE stems from the integral term of the HJBI equations; the other random measure in BSDE is introduced to link the coupling factor of the HJBI equations. We show through an extension of the dynamic programming principle that the lower value function of this game problem is the viscosity solution of the system of our coupled HJBI equations. The uniqueness of the viscosity solution is also obtained in a space of continuous functions satisfying certain growth condition. In addition, also the upper value function of the game is shown to be the solution of the associated system of coupled Issacs' type of integral-partial differential equations. As a byproduct, we obtain the existence of the value for the game problem under the well-known Isaacs' condition.

math.OC↗