Contrastive Losses and Solution Caching for Predict-and-Optimize

Mulamba, Maxime; Mandi, Jayanta; Diligenti, Michelangelo; Lombardi, Michele; Bucarey, Víctor; Guns, Tias

doi:10.24963/ijcai.2021/390

Cited by 11 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While contrastive learning of visual representations (Hjelm et al, 2019;He et al, 2020;Chen et al, 2020) and graph representations (You et al, 2020;Tong et al, 2021) have been studied extensively, it has not been explored much for COPs. Mulamba et al (2021) derive a contrastive loss for decision-focused learning to solve COPs with uncertain inputs that can be learned from historical data, where they view non-optimal solutions as negative samples. Duan et al (2022) use contrastive pre-training to learn good representations for the boolean satisfiability problem.…”

Section: Contrastive Learning For Copsmentioning

confidence: 99%

Searching Large Neighborhoods for Integer Linear Programs with Contrastive Learning

Huang¹,

Ferber²,

Tian³

et al. 2023

Preprint

View full text Add to dashboard Cite

Integer Linear Programs (ILPs) are powerful tools for modeling and solving a large number of combinatorial optimization problems. Recently, it has been shown that Large Neighborhood Search (LNS), as a heuristic algorithm, can find high quality solutions to ILPs faster than Branch and Bound. However, how to find the right heuristics to maximize the performance of LNS remains an open problem. In this paper, we propose a novel approach, CL-LNS, that delivers state-of-the-art anytime performance on several ILP benchmarks measured by metrics including the primal gap, the primal integral, survival rates and the best performing rate. Specifically, CL-LNS collects positive and negative solution samples from an expert heuristic that is slow to compute and learns a more efficient one with contrastive learning. We use graph attention networks and a richer set of features to further improve its performance.

show abstract

Section: Contrastive Learning For Copsmentioning

confidence: 99%

Searching Large Neighborhoods for Integer Linear Programs with Contrastive Learning

Huang¹,

Ferber²,

Tian³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Blackbox differentiation: For this general formulation of LSAP a wide range of work of predict-andoptimize schemes providing an end-to-end optimization ability to combinatorial optimization problems exist based on e.g., the interpolation of optimization mappings (Vlastelica et al, 2020;Sahoo et al, 2023), continuous relaxations (Amos & Kolter, 2017;Elmachtoub & Grigas, 2017;Wilder et al, 2019) or methods that bypass the calculation of gradients for the optimizer entirely using surrogate losses (Mulamba et al, 2021;Shah et al, 2022). For an in-depth review and comparison of existing approaches, the mindful reader is referred to Geng et al (2023).…”

Section: Track Construction As a Differentiable Assignment Problemmentioning

confidence: 99%

Towards Neural Charged Particle Tracking in Digital Tracking Calorimeters with Reinforcement Learning

Kortus¹,

Gauger²

2023

Preprint

View full text Add to dashboard Cite

<p>We propose a novel reconstruction scheme for reconstructing charged particles in digital tracking calorimeters using model-free reinforcement learning aiming to benefit from the rapid progress and success of neural network architectures for tracking without the dependency on simulated or manually labeled data. Here we optimize by trial-and-error a behavior policy acting as a heuristic approximation to the full combinatorial optimization problem, maximizing the physical plausibility of sampled trajectories. In modern data processing pipelines used in high energy physics experiments and related high energy physics driven applications tracking plays an essential role allowing to identify and follow charged particle trajectories traversing particle detectors. Due to the usual high multiplicity of charged particles as well as the occurring physical interactions, randomly deflecting the particles from their initial path, the reconstruction is a challenging undertaking, requiring fast, accurate and robust algorithms. Our approach works on graph-structured data, capturing possible track hypotheses through edge connections between particles in the sensitive detector layers. We demonstrate in a comprehensive study on simulated data generated for a particle detector used for proton computed tomography, the overall high potential as well as the competitiveness of our approach compared to a heuristic search algorithm and a model trained on ground truth information. Finally, we point out limitations of our approach, guiding towards a robust foundation for further development of reinforcement learning based tracking algorithms in high energy physics.</p>

show abstract

“…There is also some work that discusses alternatives to DFL that don't require coming up with custom relaxations to the optimization problems of interest. Mulamba et al [11] provide a contrastive learning-based proxy objective that doesn't require differentiating through (or even solving) the optimization problem z * (ŷ). However, this approach makes a strong assumption about the nature of the decision loss DL, while our approach can learn surrogates that are tailored to any optimization problem.…”

Section: Related Workmentioning

confidence: 99%

Decision-Focused Learning without Differentiable Optimization: Learning Locally Optimized Decision Losses

Shah¹,

Wang²,

Wilder³

et al. 2022

Preprint

View full text Add to dashboard Cite

Decision-Focused Learning (DFL) is a paradigm for tailoring a predictive model to a downstream optimisation task that uses its predictions, so that it can perform better on that specific task. The main technical challenge associated with DFL is that it requires being able to differentiate through argmin operations to work. However, these arg min optimisations are often piecewise constant and, as a result, naively differentiating through them would provide uninformative gradients. Past work has largely focused on getting around this issue by handcrafting taskspecific surrogates to the original optimisation problem that provide informative gradients when differentiated through. However, finding these surrogates can be challenging and the need to handcraft surrogates for each new task limits the usability of DFL. In addition, even after applying these relaxation techniques, there are no guarantees that the resulting surrogates are convex and, as a result, training a predictive model on them may lead to said model getting stuck in local minimas. In this paper, we provide an approach to learn faithful task-specific surrogates which (a) only requires access to a black-box oracle that can solve the optimisation problem and is thus generalizable, and (b) can be convex by construction and so can be easily optimized over. To the best of our knowledge, this is the first work on using learning to find good surrogates for DFL. We evaluate our approach on a budget allocation problem from the literature and find that our approach outperforms even the hand-crafted (non-convex) surrogate loss proposed by the original paper. Taking a step back, we hope that the generality and simplicity of our approach will help lower the barrier associated with implementing DFL-based solutions in practice. To that end, we are currently working on extending our experiments to more domains.Preprint. Under review.

show abstract

Contrastive Losses and Solution Caching for Predict-and-Optimize

Cited by 11 publications

References 0 publications

Searching Large Neighborhoods for Integer Linear Programs with Contrastive Learning

Searching Large Neighborhoods for Integer Linear Programs with Contrastive Learning

Towards Neural Charged Particle Tracking in Digital Tracking Calorimeters with Reinforcement Learning

Decision-Focused Learning without Differentiable Optimization: Learning Locally Optimized Decision Losses

Contact Info

Product

Resources

About