Searcharxiv⌕ Search

arXiv subjects

Xiaodong Luo

Publications and source records attributed to Xiaodong Luo.

At least 19 recordsLinked to original sources
arXiv2026

Equilibria for Time-inconsistent Regular-singular Control Problems

This paper studies a time-inconsistent mixed regular-singular control problem in continuous time with non-exponential discount. We seek time-consistent equilibria in an intrapersonal game framework and propose a novel definition of regular-singular equilibrium where both regular and singular controls are unified via control laws. We establish a verification theorem to provide a sufficient condition of the equilibrium. Furthermore, we propose a novel concept of mild equilibrium for which perturbations are applied separately to the regular and singular control and give the mild verification theorem. We illustrate the applicability of the theory to the effort and dividend problem and obtain an explicit equilibrium depending on the effort and dividend thresholds under a mixture of exponential discount functions and a pseudo-exponential discount function by solving the extended Hamilton--Jacobi--Bellman (HJB) system. The convexity of the value function is rigorously established, and the equilibrium conditions are verified. In addition, we establish the existence of the effort and dividend thresholds under a mixture of exponential discount functions. Numerical results reveal the impacts of exogenous parameters on the effort and dividend thresholds and the equilibrium value function, along with their economic implications.

math.OC↗
arXiv2026

A Reinforcement Learning Framework for Some Singular Stochastic Control Problems

We develop a continuous-time reinforcement learning framework for a class of singular stochastic control problems without entropy regularization. The optimal singular control is characterized as the optimal singular control law, which is a pair of regions of time and the augmented states. The goal of learning is to identify such an optimal region via the trial-and-error procedure. In this context, we generalize the existing policy evaluation theories with regular controls to learn our optimal singular control law and develop a policy improvement theorem via the region iteration. To facilitate the model-free policy iteration procedure, we further introduce the zero-order and first-order q-functions arising from singular control problems and establish the martingale characterization for the pair of q-functions together with the value function. Based on our theoretical findings, some q-learning algorithms are devised accordingly and a numerical example based on simulation experiment is presented.

math.OC↗
arXiv2026

SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD

Full-parameter post-training of trillion-parameter-scale MoE models introduces substantial system-level challenges for large-scale distributed training, including severe memory pressure, non-overlapped communication overhead, and inefficient kernel execution. While most large-scale LLM training systems are built around GPU-based clusters, this report presents an end-to-end optimization practice on the Ascend NPU SuperPOD. Using the DeepSeek-V4 model family as the target workload, we develop a hierarchical optimization framework spanning model-level parallelism, computation-communication orchestration, and low-level kernel execution. The resulting system achieves 34.22% Model FLOPs Utilization (MFU) with a 2.93x improvement over the open-source baseline recipe while maintaining training stability. Building on this optimized infrastructure, we further establish a CPT and SFT workflow for complex Operations Research (OR) tasks. We refer to the integrated framework as SLAI T-Rex. Using DeepSeek-V4-Flash, we develop OR-oriented CPT and SFT data pipelines that combine collected domain resources with solver-verified synthetic optimization documents. The resulting dataset contains 10K high-quality SFT samples spanning four task categories and three problem representations. The specialized model achieves the highest average zero-shot Pass@1 score among the evaluated models, reaching 71.81% and outperforming GPT-5.4-Mini and the base DeepSeek-V4-Flash model by 3.98 and 11.27 percentage points, respectively. Overall, this work demonstrates a full-stack pathway from efficient trillion-parameter model post-training on Ascend infra to domain-specialized Flash models for solver-grounded mathematical modeling, advancing frontier-model systems for complex reasoning.

cs.CL↗
arXiv2026

Equilibrium for regular-singular control under mean-variance criterion: A unified approach via control laws

This paper studies a class of mixed regular-singular control problems under mean-variance criteria, where the drift and diffusion are allowed to depend on both the regular control and the level of singular control. We seek time-consistent equilibrium strategies in an intrapersonal game setting and propose a novel equilibrium notion where both regular and singular controls are unified via control laws, or equivalently, mappings on the augmented state space. Under which, we derive a verification theorem and necessary conditions providing a full mathematical characterization of the equilibrium. We apply the theory to a reinsurance problem. The equilibrium solution turns out to be nontrivially coupled, where the regular control depends on the level of singular control, and the free boundary of the singular control switches dynamically in accordance with the variation of the regular control expression. The free boundary is characterized in a piecewise semi-explicit manner and is proved to be $C^{1}$ across the switching point. In the degenerate case $α_2=0$, the coupled solution reduces to the combination of two independent single-control equilibria and coincides with the limit as the parameter tends to zero.

math.OC↗
arXiv2026

Why Post-Norm Transformers Collapse: Attention Amplification and Gradient Repair Failure

Deep decoder-only Transformers often replace the original Post-Norm architecture with Pre-Norm variants because Post-Norm training is highly sensitive to warmup and learning rate under conventional initialization schemes. Although prior work has identified rank collapse and gradient vanishing as related symptoms, it remains poorly understood how causal attention creates high-similarity representations and why training dynamics fail to repair them. We give a two-stage analysis of Post-Norm rank collapse using token similarity as a scalar state variable. First, at initialization, causal attention acts approximately as a prefix-averaging operator that increases token similarity across depth, while the SwiGLU branch contributes only a smaller damping effect. Second, once training enters a high-similarity regime, growth of pre-normalization residual norms makes the RMSNorm backward factor contractive; under mild conditions, gradients to earlier layers decay geometrically. As a complementary result, we characterize the properties of a collapsed network: its best predictor is frequency distribution with relatively high loss floor, and gradients in collapsed layers vanish at frequency distribution. Experiments on 48-layer decoder-only Transformers trained on C4 dataset match the predicted initialization-time similarity growth and collapse-time gradient contraction, and show that collapsed runs stay near the predicted frequency loss. Together, these results distinguish the forward similarity amplification and backward repair incapacity in Post-Norm collapse, while also characterizing the behavior of collapsed networks.

cs.LG↗
arXiv2026

Reinforcement learning for irreversible reinsurance problems: the randomized singular control approach

This paper studies the continuous-time reinforcement learning for stochastic singular control with the application to an infinite-horizon irreversible reinsurance problem. The singular control is equivalently characterized as a pair of regions of time and the augmented states, called the singular control law. To encourage the exploration in the learning procedure, we propose a randomization method by considering an auxiliary singular control and entropy regularization. The exploratory singular control problem is formulated as a two-stage optimal control problem, in which the time-inconsistency issue arises in the outer problem. Existence of equilibrium singular control law for the time-inconsistent outer problem is rigorously established. Taking advantage of the solution structure, we utilize a proper parameterization and neural networks to devise the actor-critic reinforcement learning algorithm. In the numerical experiment, we show the superior convergence of parameter iterations based on the randomized equilibrium policy and illustrate how the exploration may advance the learning performance.

math.OC↗
arXiv2026

Equilibrium singular dividend control under ambiguity aggregation of heterogeneous discount rates

This paper studies a singular dividend control problem for a firm with heterogeneous shareholders whose discount rates follow a given distribution. The central planner aggregates expected discounted payoffs using an ambiguity aggregation function $phi$, which captures shareholder heterogeneity and ambiguity attitudes but also leads to time inconsistency. To address this issue, we seek a time-homogeneous equilibrium dividend law characterized by a partition of the state space into waiting and dividend-paying regions. We provide a rigorous mathematical characterization by proving a verification theorem and deriving necessary conditions for the equilibrium law. We then analyze barrier-type equilibria, showing non-existence for a class of aggregation functions that includes power-type and logarithmic aggregation functions, and establishing existence and uniqueness under linear and exponential aggregation. In the linear case, the bounded-rate equilibrium is shown to converge to the singular barrier-type equilibrium as the dividend rate bound tends to infinity. Numerical examples illustrate the effects of discount-rate heterogeneity and ambiguity aversion on the equilibrium barrier.

math.OC↗
arXiv2026

Time-Inconsistent Singular Control Problems with a Running Minimum Process

This paper develops a time-inconsistent and path-dependent singular control framework incorporating a running minimum process. We derive a verification theorem that characterizes equilibria under substantially weaker regularity conditions than those imposed in the existing literature, and we obtain a stronger notion of equilibrium by enlarging the class of feasible perturbations. We first establish the mathematical foundations of the framework by proving the existence and uniqueness of strong solutions to a class of Skorokhod reflection problems involving the running minimum and by characterizing admissible singular control laws. We further demonstrate the existence of an equilibrium through a dividend problem, where the running minimum leads to a highly coupled and nonlinear differential-algebraic system. For this problem, we prove the monotonicity and local concavity of the dividend boundary, thereby providing a mathematical explanation for dividend smoothing and scarring effects. Numerical simulations confirm the robustness of the equilibrium across a wide range of parameter values.

math.OC↗
arXiv2026

Finding the Balance Rate of Uncertain Signed Graphs

Signed graphs are widely used to analyze complex systems such as social, political, and biological networks. The notion of balance, a key concept of signed graphs, reflects the stability of relationships. While it has been extensively studied in deterministic graphs, real-world networks often exhibit uncertainty in their connections, which traditional approaches struggle to address. To bridge this gap, we introduce the concept of balance rate, a metric for quantifying the degree of balance in uncertain signed graphs, and prove that computing it exactly is NP-hard, motivating the need for efficient estimation methods. We propose a novel Rao-Blackwellized spanning-tree estimator that achieves near-linear time complexity per sample by leveraging graph decomposition and structural properties. We also construct asymptotically justified confidence intervals using the Delta method. Experiments on real-world datasets demonstrate the efficiency and effectiveness of our approach, enabling scalable balance analysis in uncertain signed graphs.

cs.DS↗
arXiv2026

ReMAP: Neural Reparameterization for Scalable MAP Inference in Arbitrary-Order Markov Random Fields

Scalable high-quality MAP inference in arbitrary-order Markov Random Fields (MRFs) remains challenging. Approximate message-passing methods are often efficient but can degrade on dense or high-order instances, while exact solvers such as Toulbar2 become increasingly expensive at scale. We present ReMAP, an instance-wise neural reparameterization framework that directly optimizes a differentiable relaxation of the original MRF energy. Instead of relying on supervised labels or amortized training, ReMAP treats each MRF as an independent optimization problem: a Graph Neural Network produces node-wise label distributions, and gradient-based optimization searches for a low-energy discrete solution in an over-parameterized continuous space. The method supports pairwise and arbitrary-order factors, heterogeneous label cardinalities, and efficient GPU execution, without requiring labeled solutions. We show that the relaxed objective is consistent with the discrete MAP problem and analyze how neural over-parameterization can expose low-energy optimization paths unavailable in the original discrete space. Empirically, on synthetic pairwise and high-order MRFs, UAI 2022 inference benchmarks, and real-world Physical Cell Identity (PCI) problems, ReMAP consistently outperforms approximate baselines and often finds lower-energy solutions than Toulbar2 on hard large-scale instances within practical time budgets.

cs.LG↗
arXiv2025

N2N: A Parallel Framework for Large-Scale MILP under Distributed Memory

Parallelization has emerged as a promising approach for accelerating MILP solving. However, the complexity of the branch-and-bound (B&B) framework and the numerous effective algorithm components in MILP solvers make it difficult to parallelize. In this study, a scalable parallel framework, N2N (a node-to-node framework that maps the B&B nodes to distributed computing nodes), was proposed to solve large-scale problems in a distributed memory computing environment. Both deterministic and nondeterministic modes are supported, and the framework is designed to be easily integrated with existing solvers. Regarding the deterministic mode, a novel sliding-window-based algorithm was designed and implemented to ensure that tasks are generated and solved in a deterministic order. Moreover, several advanced techniques, such as the utilization of CP search and general primal heuristics, have been developed to fully utilize distributed computing resources and capabilities of base solvers. Adaptive solving and data communication optimization were also investigated. A popular open-source MILP solver, SCIP, was integrated into N2N as the base solver, yielding N2N-SCIP. Extensive computational experiments were conducted to evaluate the performance of N2N-SCIP compared to ParaSCIP, which is a state-of-the-art distributed parallel MILP solver under the UG framework. The nondeterministic N2N-SCIP achieves speedups of 22.52 and 12.71 with 1,000 MPI processes on the Kunpeng and x86 computing clusters, which is 1.98 and 2.08 times faster than ParaSCIP, respectively. In the deterministic mode, N2N-SCIP also shows significant performance improvements over ParaSCIP across different process numbers and computing clusters. To validate the generality of N2N, HiGHS, another open-source solver, was integrated into N2N. The related results are analyzed, and the requirements of N2N on base solvers are also concluded.

cs.AI↗
arXiv2025

QuadEnhancer: Leveraging Quadratic Transformations to Enhance Deep Neural Networks

The combination of linear transformations and non-linear activation functions forms the foundation of most modern deep neural networks, enabling them to approximate highly complex functions. This paper explores the introduction of quadratic transformations to further increase nonlinearity in neural networks, with the aim of enhancing the performance of existing architectures. To reduce parameter complexity and computational complexity, we propose a lightweight quadratic enhancer that uses low-rankness, weight sharing, and sparsification techniques. For a fixed architecture, the proposed approach introduces quadratic interactions between features at every layer, while only adding negligible amounts of additional model parameters and forward computations. We conduct a set of proof-of-concept experiments for the proposed method across three tasks: image classification, text classification, and fine-tuning large-language models. In all tasks, the proposed approach demonstrates clear and substantial performance gains.

cs.LG↗
arXiv2025

Solving Quadratic Programs via Deep Unrolled Douglas-Rachford Splitting

Convex quadratic programs (QPs) are fundamental to numerous applications, including finance, engineering, and energy systems. Among the various methods for solving them, the Douglas-Rachford (DR) splitting algorithm is notable for its robust convergence properties. Concurrently, the emerging field of Learning-to-Optimize offers promising avenues for enhancing algorithmic performance, with algorithm unrolling receiving considerable attention due to its computational efficiency and interpretability. In this work, we propose an approach that unrolls a modified DR splitting algorithm to efficiently learn solutions for convex QPs. Specifically, we introduce a tailored DR splitting algorithm that replaces the computationally expensive linear system-solving step with a simplified gradient-based update, while retaining convergence guarantees. Consequently, we unroll the resulting DR splitting method and present a well-crafted neural network architecture to predict QP solutions. Our method achieves up to 50% reductions in iteration counts and 40% in solve time across benchmarks on both synthetic and real-world QP datasets, demonstrating its scalability and superior performance in enhancing computational efficiency across varying sizes.

math.OC↗
arXiv2025

Decentralized Handover Parameter Optimization with MARL for Load Balancing in 5G Networks

In cellular networks, cell handover refers to the process where a device switches from one base station to another, and this mechanism is crucial for balancing the load among different cells. Traditionally, engineers would manually adjust parameters based on experience. However, the explosive growth in the number of cells has rendered manual tuning impractical. Existing research tends to overlook critical engineering details in order to simplify handover problems. In this paper, we classify cell handover into three types, and jointly model their mutual influence. To achieve load balancing, we propose a multi-agent-reinforcement-learning (MARL)-based scheme to automatically optimize the parameters. To reduce the agent interaction costs, a distributed training is implemented based on consensus approximation of global average load, and it is shown that the approximation error is bounded. Experimental results show that our proposed scheme outperforms existing benchmarks in balancing load and improving network performance.

cs.NI↗
arXiv2025

When GNNs meet symmetry in ILPs: an orbit-based feature augmentation approach

A common characteristic in integer linear programs (ILPs) is symmetry, allowing variables to be permuted without altering the underlying problem structure. Recently, GNNs have emerged as a promising approach for solving ILPs. However, a significant challenge arises when applying GNNs to ILPs with symmetry: classic GNN architectures struggle to differentiate between symmetric variables, which limits their predictive accuracy. In this work, we investigate the properties of permutation equivariance and invariance in GNNs, particularly in relation to the inherent symmetry of ILP formulations. We reveal that the interaction between these two factors contributes to the difficulty of distinguishing between symmetric variables. To address this challenge, we explore the potential of feature augmentation and propose several guiding principles for constructing augmented features. Building on these principles, we develop an orbit-based augmentation scheme that first groups symmetric variables and then samples augmented features for each group from a discrete uniform distribution. Empirical results demonstrate that our proposed approach significantly enhances both training efficiency and predictive performance.

cs.LG↗
arXiv2025

SymILO: A Symmetry-Aware Learning Framework for Integer Linear Optimization

Integer linear programs (ILPs) are commonly employed to model diverse practical problems such as scheduling and planning. Recently, machine learning techniques have been utilized to solve ILPs. A straightforward idea is to train a model via supervised learning, with an ILP as the input and an optimal solution as the label. An ILP is symmetric if its variables can be permuted without changing the problem structure, resulting in numerous equivalent and optimal solutions. Randomly selecting an optimal solution as the label can introduce variability in the training data, which may hinder the model from learning stable patterns. In this work, we incorporate the intrinsic symmetry of ILPs and propose a novel training framework called SymILO. Specifically, we modify the learning task by introducing solution permutation along with neural network weights as learnable parameters and then design an alternating algorithm to jointly optimize the loss function. We conduct extensive experiments on ILPs involving different symmetries and the computational results demonstrate that our symmetry-aware approach significantly outperforms three existing methods -- achieving $50.3\%$, $66.5\%$, and $45.4\%$ average improvements, respectively.

math.OC↗
arXiv2024

An Efficient Unsupervised Framework for Convex Quadratic Programs via Deep Unrolling

Quadratic programs (QPs) arise in various domains such as machine learning, finance, and control. Recently, learning-enhanced primal-dual hybrid gradient (PDHG) methods have shown great potential in addressing large-scale linear programs; however, this approach has not been extended to QPs. In this work, we focus on unrolling "PDQP", a PDHG algorithm specialized for convex QPs. Specifically, we propose a neural network model called "PDQP-net" to learn optimal QP solutions. Theoretically, we demonstrate that a PDQP-net of polynomial size can align with the PDQP algorithm, returning optimal primal-dual solution pairs. We propose an unsupervised method that incorporates KKT conditions into the loss function. Unlike the standard learning-to-optimize framework that requires optimization solutions generated by solvers, our unsupervised method adjusts the network weights directly from the evaluation of the primal-dual gap. This method has two benefits over supervised learning: first, it helps generate better primal-dual gap since the primal-dual gap is in the objective function; second, it does not require solvers. We show that PDQP-net trained in this unsupervised manner can effectively approximate optimal QP solutions. Extensive numerical experiments confirm our findings, indicating that using PDQP-net predictions to warm-start PDQP can achieve up to 45% acceleration on QP instances. Moreover, it achieves 14% to 31% acceleration on out-of-distribution instances.

math.OC↗
arXiv2024

A Unified Three-State Model Framework for Analysis of Treatment Crossover in Survival Trials

We present a unified three-state model (TSM) framework for evaluating treatment effects in clinical trials in the presence of treatment crossover. Researchers have proposed diverse methodologies to estimate the treatment effect that would have hypothetically been observed if treatment crossover had not occurred. However, there is little work on understanding the connections between these different approaches from a statistical point of view. The proposed TSM framework unifies existing methods, effectively identifying potential biases, model assumptions, and inherent limitations for each method. This can guide researchers in understanding when these methods are appropriate and choosing a suitable approach for their data. The TSM framework also facilitates the creation of new methods to adjust for confounding effects from treatment crossover. To illustrate this capability, we introduce a new imputation method that falls under its scope. Through simulation experiments, we demonstrate the performance of different approaches for estimating the treatment effects. Codes for implementing the methods within the TSM framework are available at https://github.com/JasonZhao111/TSM.

stat.ME↗