| Total: 35
The classic Resource Constrained Shortest Path (RCSP) problem aims to find a cost optimal path between a pair of nodes in a network such that the resources used in the path are within a given limit. Having been studied for over a decade, RCSP has seen recent solutions that utilize heuristic-guided search to solve the constrained problem faster. Building upon the bidirectional A* search paradigm, this paper introduces a novel constrained search framework that uses efficient pruning strategies to allow for accelerated and effective RCSP search in large-scale networks. Results show that, compared to the state of the art, our enhanced framework can significantly reduce the constrained search time, achieving speed-ups of over to two orders of magnitude.
Horiyama et al. (AAAI 2024) considered the problem of generating instances with a unique minimum vertex cover under certain conditions. The Pre-assignment for Uniquification of Minimum Vertex Cover problem (shortly PAU-VC) is the problem, for given a graph G, to find a minimum set S of vertices in G such that there is a unique minimum vertex cover of G containing S. We show that PAU-VC is fixed parameter tractable parameterized by clique-width, which improves an exponential algorithm for trees given by Horiyama et al. Among natural graph classes with unbounded clique-width, we show that the problem can be solved in polynomial time on split graphs and unit interval graphs.
We initiate the study of tree structures in the context of scenario-based robust optimization. Specifically, we study Binary Search Trees (BSTs) and Huffman coding, two fundamental techniques for efficiently managing and encoding data based on a known set of frequencies of keys. Given a number of distinct scenarios, each defined by a frequency distribution over the keys, our objective is to compute a single tree of best-possible performance, relative to any scenario. We consider, as performance metrics, the competitive ratio, which compares multiplicatively the cost of the solution to the tree of least cost among all scenarios, as well as the regret, which induces a similar, but additive comparison. For BSTs, we show that the problem is NP-hard across both metrics. We also obtain an optimal competitive ratio that is logarithmic in the number of scenarios. For Huffman Trees, we likewise prove NP-hardness, and we present an algorithm with logarithmic regret, which we prove to be near-optimal by showing a corresponding lower bound. Last, we give a polynomial-time algorithm for computing Pareto-optimal BSTs with respect to their regret, assuming scenarios defined by uniform distributions over the keys. This setting captures, in particular, the first study of fairness in the context of data structures. We provide an experimental evaluation of all algorithms. To this end, we also provide mixed integer linear program formulation for computing optimal trees.
The clique partitioning problem (CPP) aims to find a partition of vertices of a complete graph in order to maximize the sum of edge weights within each partition (clique), which has been proven to be NP-hard and has wide real-world applications. In this paper, we propose an elite-guided weighted simulated annealing algorithm called EWSA to solve the CPP. First, EWSA employs two specific configurations and alternates between them via an oscillation strategy, which balances the exploitation and exploration of the search. Second, a weighting strategy is introduced to improve the scoring function in traditional simulated annealing, which is able to guide the search to explore diverse solutions. Finally, a partition restriction strategy is adopted to reduce search space and increase the search efficiency. Experiments on 255 instances demonstrate the competitiveness of EWSA. For 130 open instances, EWSA discovers new upper bounds in 32 cases and matches the best known results for the others. For the remaining 125 closed instances, EWSA achieves the best known objective values within a short computational time.
Binary Neural Networks (BNNs) have garnered significant attention due to their immense potential for deployment on edge devices. However, the non-differentiability of the quantization function poses a challenge for the optimization of BNNs, as its derivative cannot be backpropagated. To address this issue, hypernetwork based methods, which utilize neural networks to learn the gradients of non-differentiable quantization functions, have emerged as a promising approach due to their adaptive learning capabilities to reduce estimation errors. However, existing hypernetwork based methods typically rely solely on current gradient information, neglecting the influence of historical gradients. This oversight can lead to accumulated gradient errors when calculating gradient momentum during optimization. To incorporate historical gradient information, we design a Historical Gradient Storage (HGS) module, which models the historical gradient sequence to generate the first-order momentum required for optimization. To further enhance gradient generation in hypernetworks, we propose a Fast and Slow Gradient Generation (FSG) method. Additionally, to produce more precise gradients, we introduce Layer Recognition Embeddings (LRE) into the hypernetwork, facilitating the generation of layer-specific fine gradients. Extensive comparative experiments on the CIFAR-10 and CIFAR-100 datasets demonstrate that our method achieves faster convergence and lower loss values, outperforming existing baselines.
Somatic Contiguous Hypermutations (CHM) are a popular variation operator used in artificial immune systems for optimisation tasks. Theoretical studies have shown that CHM operators can lead to considerable speed-ups in the expected optimisation time compared to the traditional standard bit mutation (SBM) operators used in evolutionary computation for both single-objective and multi-objective problems where it is advantageous to mutate large contiguous areas of the genotype representing the candidate solutions. These speed-ups can make the difference between polynomial and exponential runtimes, but come at the expense of the CHM operator being considerably slower than the SBM operator in easy hillclimbing phases of the optimisation process, when small areas of the genotype have to be mutated for progress to be made. In this paper we present a Fast CHM operator that is asymptotically just as fast as traditional SBM for hillclimbing yet maintains the efficacy of the standard CHM operator when large jumps in the search space are required to make progress efficiently. We demonstrate such efficacy on all applications were CHM has been previously studied in the literature.
Automatic Heuristic Design (AHD) is an active research area due to its utility in solving complex search and NP-hard combinatorial optimization problems in the real world. The recent advancements in Large Language Models (LLMs) introduce new possibilities by coupling LLMs with evolutionary computation to automatically generate heuristics, known as LLM-based Evolutionary Program Search (LLM-EPS). While previous LLM-EPS studies obtained great performance on various tasks, there is still a gap in understanding the properties of heuristic search spaces and achieving a balance between exploration and exploitation, which is a critical factor in large heuristic search spaces. In this study, we address this gap by proposing two diversity measurement metrics and perform an analysis on previous LLM-EPS approaches, including FunSearch, EoH, and ReEvo. Results on black-box AHD problems reveal that while EoH demonstrates higher diversity than FunSearch and ReEvo, its objective score is unstable. Conversely, ReEvo's reflection mechanism yields good objective scores but fails to optimize diversity effectively. With this finding in mind, we introduce HSEvo, an adaptive LLM-EPS framework that maintains a balance between diversity and convergence with a harmony search algorithm. Through experimentation, we find that HSEvo achieved high diversity indices and good objective scores while remaining cost-effective. These results underscore the importance of balancing exploration and exploitation and understanding heuristic search spaces in designing frameworks in LLM-EPS.
A graph G is c-closed if every two vertices with at least c common neighbors are adjacent to each other. This definition is an abstraction of the triadic closure property exhibited by many real-world social networks, namely, friends of friends tend to be friends themselves. Social networks, however, are often temporal rather than static---the connections change over a period of time. And hence temporal graphs, rather than static graphs, are often better suited to model social networks. Motivated by this, we introduce a definition of temporal c-closed graphs, in which if two vertices u and v have at least c common neighbors during a short interval of time, then u and v are adjacent to each other around that time. Our pilot experiments show that several real-world temporal networks are c-closed for rather small values of c. We also study the computational problems of enumerating maximal cliques and other dense subgraphs in temporal c-closed graphs. A clique in a temporal graph is a subgraph that lasts for a certain period of time, during which every possible edge in the subgraph becomes active often enough; other dense subgraphs are defined similarly. We bound the number of such maximal dense subgraphs in a temporal c-closed graph that evolves slowly, and thus show that the corresponding enumeration problems admit efficient algorithms; by slow evolution, we mean that between consecutive time-steps, the local change in adjacencies remains small. Our work also adds to a growing body of literature on defining suitable structural parameters for temporal graphs that can be leveraged to design efficient algorithms.
We present Learn2Aggregate, a machine learning (ML) framework for optimizing the generation of Chvatal-Gomory (CG) cuts in mixed integer linear programming (MILP). The framework trains a graph neural network to classify useful constraints for aggregation in CG cut generation. The ML-driven CG separator selectively focuses on a small set of impactful constraints, improving runtimes without compromising the strength of the generated cuts. Key to our approach is the formulation of a constraint classification task which favours sparse aggregation of constraints, consistent with empirical findings. This, in conjunction with a careful constraint labeling scheme and a hybrid of deep learning and feature engineering, results in enhanced CG cut generation across five diverse MILP benchmarks. On the largest test sets, our method closes roughly twice as much of the integrality gap as the standard CG method while running 40% faster. This performance improvement is due to our method eliminating 75% of the constraints prior to aggregation.
Randomized search heuristics have been applied successfully to a plethora of problems. This success is complemented by a large body of theoretical results. Unfortunately, the vast majority of these results regard problems with binary or continuous decision variables -- the theoretical analysis of randomized search heuristics for unbounded integer domains is almost nonexistent. To resolve this shortcoming, we start the runtime analysis of multi-objective evolutionary algorithms, which are among the most successful randomized search heuristics, for unbounded integer search spaces. We analyze single- and full-dimensional mutation operators with three different mutation strengths, namely changes by plus/minus one (unit strength), random changes following a law with exponential tails, and random changes following a power-law. The performance guarantees we prove on a recently proposed natural benchmark problem suggest that unit mutation strengths can be slow when the initial solutions are far from the Pareto front. When setting the expected change right (depending on the benchmark parameter and the distance of the initial solutions), the mutation strength with exponential tails yields the best runtime guarantees in our results -- however, with a wrong choice of this expectation, the performance guarantees quickly become highly uninteresting. With power-law mutation, which is an essentially parameter-less mutation operator, we obtain good results uniformly over all problem parameters and starting points. We complement our mathematical findings with experimental results that suggest that our bounds are not always tight. Most prominently, our experiments indicate that power-law mutation outperforms the one with exponential tails even when the latter uses a near-optimal parametrization. Hence, we suggest to favor power-law mutation for unknown problems in integer spaces.
The non-dominated sorting genetic algorithm II (NSGA-II) is the most popular multi-objective optimization heuristic. Recent mathematical runtime analyses have detected two shortcomings in discrete search spaces, namely, that the NSGA-II has difficulties with more than two objectives and that it is very sensitive to the choice of the population size. To overcome these difficulties, we analyze a simple tie-breaking rule in the selection of the next population. Similar rules have been proposed before, but have found only little acceptance. We prove the effectiveness of our tie-breaking rule via mathematical runtime analyses on the classic OneMinMax, LeadingOnesTrailingZeros, and OneJumpZeroJump benchmarks. We prove that this modified NSGA-II can optimize the three benchmarks efficiently also for many objectives, in contrast to the exponential lower runtime bound previously shown for OneMinMax with three or more objectives. For the bi-objective problems, we show runtime guarantees that do not increase when moderately increasing the population size over the minimum admissible size. For example, for the OneJumpZeroJump problem with representation length n and gap parameter k, we show a runtime guarantee of O(max {n^(k + 1), N n}) function evaluations when the population size is at least four times the size of the Pareto front. For population sizes larger than the minimal choice N = Θ(n), this result improves considerably over the Θ(N n^k) runtime of the classic NSGA-II.
We consider the Online Facility Location (OFL) problem in the framework of learning-augmented online algorithms. In Online Facility Location (OFL), demands arrive one-by-one in a metric space and must be (irrevocably) assigned to an open facility upon arrival, without any knowledge about future demands. We focus on uniform facility opening costs and present an online algorithm for OFL that exploits potentially imperfect predictions on the locations of the optimal facilities. We prove that the competitive ratio decreases from sublogarithmic in the number n of demands to constant as the so-called η1 error, i.e., the sum of distances of the predicted locations to the optimal facility locations, decreases towards zero. E.g., our analysis implies that if for some ε > 0, η1 = OPT / n^ε, where OPT is the cost of the optimal solution, the competitive ratio is O(1/ε). We complement our analysis with a matching lower bound establishing that the dependence of the algorithm's competitive ratio on the η1 error is optimal, up to constant factors.
Recent advances in Meta-learning for Black-Box Optimization (MetaBBO) have shown the potential of using neural networks to dynamically configure evolutionary algorithms (EAs), enhancing their performance and adaptability across various BBO instances. However, they are often tailored to a specific EA, which limits their generalizability and necessitates retraining or redesigns for different EAs and optimization problems. To address this limitation, we introduce ConfigX, a new paradigm of the MetaBBO framework that is capable of learning a universal configuration agent (model) for boosting diverse EAs. To achieve so, our ConfigX first leverages a novel modularization system that enables the flexible combination of various optimization sub-modules to generate diverse EAs during training. Additionally, we propose a Transformer-based neural network to meta-learn a universal configuration policy through multitask reinforcement learning across a designed joint optimization task space. Extensive experiments verify that, our ConfigX, after large-scale pre-training, achieves robust zero-shot generalization to unseen tasks and outperforms state-of-the-art baselines. Moreover, ConfigX exhibits strong lifelong learning capabilities, allowing efficient adaptation to new tasks through fine-tuning. Our proposed ConfigX represents a significant step toward an automatic, all-purpose configuration agent for EAs.
In bounded-suboptimal heuristic search, the aim is to find a solution path within a given bound as quickly as possible, which is crucial when computational resources are limited. Recent research has demonstrated Weighted A* variants such as XDP that find bounded suboptimal solutions without needing to perform state re-expansions; they work by shifting where the suboptimality in the search is allowed. However, the suboptimality distribution is fixed before the search begins. This paper introduces Dynamic Suboptimality Weighted A* (DSWA*), a search framework that allows suboptimality to be dynamically distributed at runtime, based on the properties of the search. Experiments show that dynamic policies can consistently outperform existing algorithms across a diverse set of domains, particularly those with dynamic costs.
Bayesian Optimization (BO) is a widely-used method for optimizing expensive-to-evaluate black-box functions. Traditional BO assumes that the learner has full control over all query variables without additional constraints. However, in many real-world scenarios, controlling certain query variables may incur costs. Therefore, the learner needs to balance the selection of informative subsets for targeted learning against leaving some variables to be randomly sampled to minimize costs. This problem is known as Bayesian Optimization with cost-varying variable subsets (BOCVS). While the goal of BOCVS is to identify the optimal solution with minimal cost, previous works have only guaranteed finding the optimal solution without considering the total costs incurred. Moreover, these works assume precise knowledge of the cost for each subset, which is often unrealistic. In this paper, we propose a novel algorithm for the extension of the BOCVS problem with random and unknown costs that separates the process into exploration and exploitation phases. The exploration phase will filter out low-quality variable subsets, while the exploitation phase will leverage high-quality ones. Furthermore, we theoretically demonstrate that our algorithm achieves a sub-linear rate in both quality regret and cost regret, addressing the objective of the BOCVS problem more effectively than previous analyses. Finally, we show that our proposed algorithm outperforms comparable baselines across a wide range of benchmarks.
Multi-objective Optimization Problems (MOPs) where objectives have different levels of importance in decision-making, known as Mixed Pareto-Lexicographic MOPs with Priority Levels (PL-MPL-MOPs), are increasingly prevalent in real-world applications. General-purpose Multi-Objective Evolutionary Algorithms (MOEAs) that treat all objectives equally not only increase the workload of decision-making but also suffer from computational inefficiencies due to the necessity of generating many additional solutions. Conversely, strictly adhering to Priority Levels (PLs) during optimization can easily result in premature convergence within some PLs. To address this issue, we suggest an effective Balanced Adaptive Subspace Collaboration (BASC) method in this paper. Specifically, this method decomposes the search space into sub-fronts based on PLs and utilizes a sampling mechanism that operates exclusively within subspaces formed by sub-fronts at the same PL to generate new solutions, thereby emphasizing the exploitation of individual PLs. Furthermore, a set of parameters is employed to control the strictness of adherence to each PL, with these parameters adaptively adjusted to balance exploration across different PLs. The two mechanisms are then collaboratively integrated into MOEAs. Comprehensive experimental studies on benchmark problems and a set of complex job-shop scheduling problems in semiconductor manufacturing demonstrate the competitiveness of the proposed method over existing methods.
In this paper, we propose a class of faster double adaptive gradient methods to solve nonconvex finite-sum optimization problems possibly with nonsmooth regularization by simultaneously using adaptive learning rate and adaptive mini-batch size. Specifically, we first propose a double adaptive stochastic gradient method (i.e., 2AdaSGD), and prove that our 2AdaSGD obtains a low stochastic first-order oracle (SFO) complexity for finding a stationary solution under the population smoothness condition. Furthermore, we propose a variance reduced double adaptive stochastic gradient method (i.e., 2AdaSPIDER), and prove that our 2AdaSPIDER obtains an optimal SFO complexity under the average smoothness condition, which is lower than the SFO complexity of the existing double adaptive gradient algorithms. In particular, we introduce a new stochastic gradient mapping to adaptively adjust mini-batch size in our stochastic gradient methods. We conduct some numerical experiments to verify efficiency of our proposed methods.
Batch Bayesian Optimisation (BBO) has emerged as a potent approach for optimising expensive black-box functions. Central to BBO is the issue of selecting a number of solutions at the same time through a batch method, in the hope for them to represent good, yet different, trade-offs between exploitation and exploration. To address this issue, one of the recent advancements has leveraged multi-objective optimisation to simultaneously consider several acquisition functions (e.g., PI, EI, and LCB), allowing them to complement each other. However, acquisition functions may behave similarly (since they all aim for a good balance between exploitation and exploration), restricting the search on different promising areas. In this paper, we attempt to address the above issue. We directly treat exploitation (reflected by quality, i.e., the posterior mean) and exploration (reflected by uncertainty) as two objectives. When selecting trade-off solutions between the two objectives, we consider a dynamically updated Pareto front where the uncertainty changes once a solution is selected, thereby allowing exploration on different promising areas. Through an extensive experiment study, we show the effectiveness of the proposed method in comparison with state-of-the-arts in the area.
Discovering elements of a hidden set, also known as Group Testing (GT), is a well-established area in which one party tries to discover elements hidden by the other party by asking queries and analyzing feedback. The feedback is a function of the intersection of the query with the hidden set - in our case, it is a classical double-threshold function, which returns i if the intersection is a singleton i and "null" otherwise (i.e., when the intersection is empty or of size at least 2). In this work, we enhance GT by two features. First, we introduce a local feedback framework to this problem: each hidden element is an "autonomous" element and can analyze feedback itself, but only for the queries to which it belongs. The goal is to design a deterministic non-adaptive sequence of queries that enables each non-hidden element to learn about all other hidden elements. We show that, surprisingly, this task requires substantially more queries than the classic group testing -- by proving a super-cubic (in terms of the number of hidden elements) lower bound and by constructing a specific query sequence of slightly longer length. Such a query system is also an extension of a well-known superimposed code, in a way that the decoding can be done only by the owners of the codewords. Second, we extend the results to the model where elements may belong to certain clusters and retrieving them could be done only via queries avoiding elements from "interfering" clusters. The main challenge is in not knowing which interfering clusters are non-empty (and thus, need to be avoided) and how to speed up the retrieval process by asking queries across many clusters. Our algorithms can be generalized to other feedback functions, to adversarial/stochastic fault-prone scenarios, implemented in a distributed setting and applied to the information theory and codes.
In recent years the understanding of optimal bidirectional heuristic search (BiHS) has progressed significantly. Yet, Bi-HS is relatively unexplored in unbounded suboptimal search. Front-to-end (F2E) and front-to-front (F2F) bidirectional search have been used in optimal algorithms, but adapting them for unbounded suboptimal search remains an open challenge. We introduce a framework for suboptimal BiHS, called anchor search, and use it to derive a parameterized family of algorithms. Because our new algorithms need F2F heuristic evaluations, we propose using pattern databases (PDBs) as differential heuristics (DHs) to construct F2F heuristics. Our experiments evaluate three anchor search instances across diverse domains, outperforming existing methods, particularly as the search scales.
The maximin optimisation problem, inspired by Von Neumann’s work (von Neumann 1928) and widely applied in adversarial optimisation, has become a key research area in machine learning. Gradient Descent Ascent (GDA) is a common method for solving these problems but requires the pay-off function to be differentiable, making it unsuitable for discrete or binary functions that often occur in game-theoretical scenarios. Co-evolutionary algorithms (CoEAs), which are derivative-free, offer an alternative to these problems. However, the theoretical understanding of CoEAs is still limited. This paper provides the first rigorous runtime analysis of CoEAs with pairwise dominance on binary two-player zero-sum games (or maximin problems), specifically focusing on the DIAGONAL game. The mathematical analysis rigorously shows that the PDCoEA can efficiently find the optimum in polynomial runtime with high probability under low mutation rates and large population sizes. Empirical evidence also identifies an error threshold where higher mutation rates lead to inefficiency. In contrast, single-pair-individual algorithms, i.e., RLS-PD and (1+1)-CoEAs, fail to find the optimum in polynomial time. These findings highlight the usefulness of pairwise dominance, low mutation rates, and large populations in maintaining a “co-evolutionary arms race”.
Multi-objective Bayesian optimization (MOBO) has shown promising performance on various expensive multi-objective optimization problems (EMOPs). However, effectively modeling complex distributions of the Pareto optimal solutions is difficult with limited function evaluations. Existing Pareto set learning algorithms may exhibit considerable instability in such expensive scenarios, leading to significant deviations between the obtained solution set and the Pareto set (PS). In this paper, we propose a novel Composite Diffusion Model based Pareto Set Learning algorithm (CDM-PSL) for expensive MOBO. CDM-PSL includes both unconditional and conditional diffusion model for generating high-quality samples efficiently. Besides, we introduce a weighting method based on information entropy to balance different objectives. This method is integrated with a guiding strategy to appropriately balancing different objectives during the optimization process. Experimental results on both synthetic and real-world problems demonstrates that CDM-PSL attains superior performance compared with state-of-the-art MOBO algorithms.
Recent research demonstrates the effectiveness of using pre-trained language models for legal case retrieval. Most of the existing works focus on improving the representation ability for the contextualized embedding of the [CLS] token and calculate relevance using textual semantic similarity. However, in the legal domain, textual semantic similarity does not always imply that the cases are relevant enough. Instead, relevance in legal cases primarily depends on the similarity of key facts that impact the final judgment. Without proper treatments, the discriminative ability of learned representations could be limited since legal cases are lengthy and contain numerous non-key facts. To this end, we introduce DELTA, a discriminative model designed for legal case retrieval. The basic idea involves pinpointing key facts in legal cases and pulling the contextualized embedding of the [CLS] token closer to the key facts while pushing away from the non-key facts, which can warm up the case embedding space in an unsupervised manner. To be specific, this study brings the word alignment mechanism to the contextual masked auto-encoder. First, we leverage shallow decoders to create information bottlenecks, aiming to enhance the representation ability. Second, we employ the deep decoder to enable ``translation'' between different structures, with the goal of pinpointing key facts to enhance discriminative ability. Comprehensive experiments conducted on publicly available legal benchmarks show that our approach can outperform existing state-of-the-art methods in legal case retrieval. It provides a new perspective on the in-depth understanding and processing of legal case documents.
Mathematical formulas are the language of communication between humans and nature. Discovering latent formulas from observed data is an important challenge in artificial intelligence, commonly known as symbolic regression(SR). The current mainstream SR algorithms regard SR as a combinatorial optimization problem and use Genetic Programming (GP) or Reinforcement Learning (RL) to solve the SR problem. These methods perform well on simple problems, but poorly on slightly more complex tasks. In addition, this class of algorithms ignores an important aspect: in SR tasks, symbols have explicit numerical meaning. So can we take full advantage of this important property and try to solve the SR problem with more efficient numerical optimization methods? Extrapolation and Learning Equation (EQL) replaces activation functions in neural networks with basic symbols and sparsifies connections to derive a simplified expression from a large network. However, EQL's fixed network structure can't adapt to the complexity of different tasks, often resulting in redundancy or insufficient, limiting its effectiveness. Based on the above analysis, we propose MetaSymNet, a tree-like network that employs the PANGU meta-function as its activation function. PANGU meta-function can evolve into various candidate functions during training. The network structure can also be adaptively adjusted according to different tasks. Then the symbol network evolves into a concise, interpretable mathematical expression. To evaluate the performance of MetaSymNet and five baseline algorithms, we conducted experiments across more than ten datasets, including SRBench. The experimental results show that MetaSymNet has achieved relatively excellent results on various evaluation metrics.
By exploiting the correlation between the structure and the solution of Mixed-Integer Linear Programming (MILP), Machine Learning (ML) has become a promising method for solving large-scale MILP problems. Existing ML-based MILP solvers mainly focus on end-to-end solution learning, which suffers from the scalability issue due to the high dimensionality of the solution space. Instead of directly learning the optimal solution, this paper aims to learn a reduced and equivalent model of the original MILP as an intermediate step. The reduced model often corresponds to interpretable operations and is much simpler, enabling us to solve large-scale MILP problems much faster than existing commercial solvers. However, current approaches rely only on the optimal reduced model, overlooking the significant preference information of all reduced models. To address this issue, this paper proposes a preference-based model reduction learning method, which considers the relative performance (i.e., objective cost and constraint feasibility) of all reduced models on each MILP instance as preferences. We also introduce an attention mechanism to capture and represent preference information, which helps improve the performance of model reduction learning tasks. Moreover, we propose a SetCover based pruning method to control the number of reduced models (i.e., labels), thereby simplifying the learning process. Evaluation on real-world MILP problems shows that 1) compared to the state-of-the-art model reduction ML methods, our method obtains nearly 20% improvement on solution accuracy, and 2) compared to the commercial solver Gurobi, two to four orders of magnitude speedups are achieved.