arXiv subjects

Linxin Yang

Publications and source records attributed to Linxin Yang.

15 recordsLinked to original sources

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

From Sequential to Parallel: Reformulating Dynamic Programming as GPU Kernels for Large-Scale Stochastic Combinatorial Optimization

A major bottleneck in scenario-based Sample Average Approximation (SAA) for stochastic programming (SP) is the cost of solving an exact second-stage problem for every scenario, especially when each scenario contains an NP-hard combinatorial structure. This has led much of the SP literature to restrict the second stage to linear or simplified models. We develop a GPU-based framework that makes structured integer recourse operators tractable at scale. The key innovation is a set of hardware-aware, scenario-batched GPU kernels that expose parallelism across scenarios, dynamic-programming (DP) layers, and route or action options, enabling Bellman updates to be executed in a single pass over more than 1,000,000 realizations. We evaluate the approach in two representative SP settings: a vectorized split operator for stochastic vehicle routing and a DP for inventory reinsertion. Implementation scales nearly linearly in the number of scenarios and achieves a one-two to four-five orders of magnitude speedup, allowing far larger scenario sets and reliably stronger first-stage decisions. The computational leverage directly improves decision quality: much larger scenario sets and many more first-stage candidates can be evaluated within fixed time budgets, consistently yielding stronger SAA solutions. Our results show that structured integer recourse operators are tractable at scales previously considered impossible, providing a practical path to large-scale, realistic stochastic discrete optimization.

math.OC↗

arXiv2025

GPU-based Split algorithm for Large-Scale CVRPSD

Dynamic programming (DP) is a cornerstone of combinatorial optimization, yet its inherently sequential structure has long limited its scalability in scenario-based stochastic programming (SP). This paper introduces a GPU-accelerated framework that reformulates a broad class of forward DP recursions as batched min-plus matrix-vector products over layered DAGs, collapsing actions into masked state-to-state transitions that map seamlessly to GPU kernels. Using this reformulation, our approach takes advantage of massive parallelism across both scenarios and transitions, enabling the simultaneous evaluation of \emph{over one million uncertainty realizations} in a single GPU pass -- a scale far beyond the reach of existing methods. We instantiate the framework in two canonical applications: the capacitated vehicle routing problem with stochastic demand and a dynamic stochastic inventory routing problem. In both cases, DP subroutines traditionally considered sequential are redesigned to harness two- or three-dimensional GPU parallelism. Experiments demonstrate near-linear scaling in the number of scenarios and yield one to three orders of magnitude speedups over multithreaded CPU baselines, resulting in tighter SAA estimates and significantly stronger first-stage decisions under fixed time budgets. Beyond these applications, our work establishes a general-purpose recipe for transforming classical DP routines into high-throughput GPU primitives, substantially expanding the computational frontier of stochastic discrete optimization to the million-scenario scale.

math.OC↗

arXiv2025

Successive Fixing for Large-Scale SCUC Using First-Order Methods

Security-Constrained Unit Commitment is a fundamental optimization problem in power systems operations. The primary computational bottleneck arises from the need to solve large-scale Linear Programming (LP) relaxations within branch-and-cut. Conventional simplex and barrier methods become computationally prohibitive at this scale due to their reliance on expensive matrix factorizations. While matrix-free first-order methods present a promising alternative, their tendency to converge to non-vertex solutions renders them incompatible with standard branch-and-cut procedures. To bridge this gap, we propose a successive fixing framework that leverages a customized GPU-accelerated first-order LP solver to guide a logic-driven variable-fixing strategy. Each iteration produces a reduced Mixed-Integer Linear Programming (MILP) problem, which is subsequently tightened via presolving. This iterative cycle of relaxation, fixing, and presolving progressively reduces problem complexity, producing a highly tractable final MILP model. When evaluated on public benchmarks exceeding 13,000 buses, our approach achieves a tenfold speedup over state-of-the-art methods without compromising solution quality.

math.OC↗

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

Feature Augmentation of GNNs for ILPs: Local Uniqueness Suffices

Integer Linear Programs (ILPs) are central to real-world optimizations but notoriously difficult to solve. Learning to Optimize (L2O) has emerged as a promising paradigm, with Graph Neural Networks (GNNs) serving as the standard backbone. However, standard anonymous GNNs are limited in expressiveness for ILPs, and the common enhancement of augmenting nodes with globally unique identifiers (UIDs) typically introduces spurious correlations that severely harm generalization. To address this tradeoff, we propose a parsimonious Local-UID scheme based on d-hop uniqueness coloring, which ensures identifiers are unique only within each node's d-hop neighborhood. Building on this scheme, we introduce ColorGNN, which incorporates color information via color-conditioned embeddings, and ColorUID, a lightweight feature-level variant. We prove that for d-layer networks, Local-UIDs achieve the expressive power of Global-UIDs while offering stronger generalization. Extensive experiments show that our approach yields substantial and robust gains across ILP benchmarks.

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

Relax-and-Cut for Temporal SCUC Decomposition

The Security-Constrained Unit Commitment (SCUC) problem presents formidable computational challenges due to its combinatorial complexity, large-scale network dimensions, and numerous security constraints. While conventional temporal decomposition methods achieve computational tractability through fixed short-term time windows, this limited look-ahead capability often results in suboptimal, myopic solutions. We propose an innovative relax-and-cut framework that alleviates these limitations through two key innovations. First, our enhanced temporal decomposition strategy maintains integer variables for immediate unit commitment decisions while relaxing integrality constraints for future time periods, thereby extending the optimization horizon without compromising tractability. Second, we develop a dynamic cutting-plane mechanism that selectively incorporates N-1 contingency constraints during the branch-and-cut process, avoiding the computational burden of complete upfront enumeration. The framework optionally employs a Relaxation-Induced Neighborhood Search procedure for additional solution refinement when computational resources permit. Comprehensive numerical experiments demonstrate the effectiveness of our approach on large-scale systems up to 13,000 buses. The proposed method can achieve optimality gaps below 1% while requiring only 20% of the computation time of monolithic Gurobi solutions. Compared to existing decomposition approaches, our framework provides superior performance, simultaneously reducing primal gaps by 60% and doubling solution speed. These significant improvements make our method particularly well-suited for practical SCUC implementations where both solution quality and computational efficiency are crucial.

math.OC↗

arXiv2025

A Learning-Based Inexact ADMM for Solving Quadratic Programs

Convex quadratic programs (QPs) constitute a fundamental computational primitive across diverse domains including financial optimization, control systems, and machine learning. The alternating direction method of multipliers (ADMM) has emerged as a preferred first-order approach due to its iteration efficiency - exemplified by the state-of-the-art OSQP solver. Machine learning-enhanced optimization algorithms have recently demonstrated significant success in speeding up the solving process. This work introduces a neural-accelerated ADMM variant that replaces exact subproblem solutions with learned approximations through a parameter-efficient Long Short-Term Memory (LSTM) network. We derive convergence guarantees within the inexact ADMM formalism, establishing that our learning-augmented method maintains primal-dual convergence while satisfying residual thresholds. Extensive experimental results demonstrate that our approach achieves superior solution accuracy compared to existing learning-based methods while delivering significant computational speedups of up to $7\times$, $28\times$, and $22\times$ over Gurobi, SCS, and OSQP, respectively. Furthermore, the proposed method outperforms other learning-to-optimize methods in terms of solution quality. Detailed performance analysis confirms near-perfect compliance with the theoretical assumptions, consequently ensuring algorithm convergence.

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

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

PDHG-Unrolled Learning-to-Optimize Method for Large-Scale Linear Programming

Solving large-scale linear programming (LP) problems is an important task in various areas such as communication networks, power systems, finance and logistics. Recently, two distinct approaches have emerged to expedite LP solving: (i) First-order methods (FOMs); (ii) Learning to optimize (L2O). In this work, we propose an FOM-unrolled neural network (NN) called PDHG-Net, and propose a two-stage L2O method to solve large-scale LP problems. The new architecture PDHG-Net is designed by unrolling the recently emerged PDHG method into a neural network, combined with channel-expansion techniques borrowed from graph neural networks. We prove that the proposed PDHG-Net can recover PDHG algorithm, thus can approximate optimal solutions of LP instances with a polynomial number of neurons. We propose a two-stage inference approach: first use PDHG-Net to generate an approximate solution, and then apply PDHG algorithm to further improve the solution. Experiments show that our approach can significantly accelerate LP solving, achieving up to a 3$\times$ speedup compared to FOMs for large-scale LP problems.

cs.LG↗

arXiv2023

A GNN-Guided Predict-and-Search Framework for Mixed-Integer Linear Programming

Mixed-integer linear programming (MILP) is widely employed for modeling combinatorial optimization problems. In practice, similar MILP instances with only coefficient variations are routinely solved, and machine learning (ML) algorithms are capable of capturing common patterns across these MILP instances. In this work, we combine ML with optimization and propose a novel predict-and-search framework for efficiently identifying high-quality feasible solutions. Specifically, we first utilize graph neural networks to predict the marginal probability of each variable, and then search for the best feasible solution within a properly defined ball around the predicted solution. We conduct extensive experiments on public datasets, and computational results demonstrate that our proposed framework achieves 51.1% and 9.9% performance improvements to MILP solvers SCIP and Gurobi on primal gaps, respectively.

math.OC↗

arXiv2022

The Machine Learning for Combinatorial Optimization Competition (ML4CO): Results and Insights

Combinatorial optimization is a well-established area in operations research and computer science. Until recently, its methods have focused on solving problem instances in isolation, ignoring that they often stem from related data distributions in practice. However, recent years have seen a surge of interest in using machine learning as a new approach for solving combinatorial problems, either directly as solvers or by enhancing exact solvers. Based on this context, the ML4CO aims at improving state-of-the-art combinatorial optimization solvers by replacing key heuristic components. The competition featured three challenging tasks: finding the best feasible solution, producing the tightest optimality certificate, and giving an appropriate solver configuration. Three realistic datasets were considered: balanced item placement, workload apportionment, and maritime inventory routing. This last dataset was kept anonymous for the contestants.

cs.LG↗

arXiv2022

Efficient primal heuristics for mixed-integer linear programs

This paper is a short report about our work for the primal task in the Machine Learning for Combinatorial Optimization NeurIPS 2021 Competition. For each dataset of our interest in the competition, we propose customized primal heuristic methods to efficiently identify high-quality feasible solutions. The computational studies demonstrate the superiority of our proposed approaches over the competitors'.

math.OC↗