Landscape Surrogate: Learning Decision Losses for Mathematical Optimization Under Partial Information

TL;DR

This paper introduces a learnable landscape surrogate model that accelerates optimization under partial information by replacing slow, sparse-gradient solvers with a neural network-based surrogate, improving efficiency and generalization.

Contribution

The paper proposes a novel neural network-based landscape surrogate that replaces traditional solvers, enabling faster training and deployment in partial information optimization problems.

Findings

Achieves comparable or better objective values than state-of-the-art methods.

Reduces the number of calls to the original optimizer $ extbf{g}$.

Outperforms existing methods on high-dimensional, computationally expensive problems.

Abstract

Recent works in learning-integrated optimization have shown promise in settings where the optimization problem is only partially observed or where general-purpose optimizers perform poorly without expert tuning. By learning an optimizer $\mathbf{g}$ to tackle these challenging problems with $f$ as the objective, the optimization process can be substantially accelerated by leveraging past experience. The optimizer can be trained with supervision from known optimal solutions or implicitly by optimizing the compound function $f\circ \mathbf{g}$. The implicit approach may not require optimal solutions as labels and is capable of handling problem uncertainty; however, it is slow to train and deploy due to frequent calls to optimizer $\mathbf{g}$ during both training and testing. The training is further challenged by sparse gradients of $\mathbf{g}$, especially for combinatorial solvers. To address these challenges, we propose using a smooth and learnable Landscape Surrogate $M$ as a replacement for $f\circ \mathbf{g}$. This surrogate, learnable by neural networks, can be computed faster than the solver $\mathbf{g}$, provides dense and smooth gradients during training, can generalize to unseen optimization problems, and is efficiently learned via alternating optimization. We test our approach on both synthetic problems, including shortest path and multidimensional knapsack, and real-world problems such as portfolio optimization, achieving comparable or superior objective values compared to state-of-the-art baselines while reducing the number of calls to $\mathbf{g}$. Notably, our approach outperforms existing methods for computationally expensive high-dimensional problems.