End-to-End Learning for Stochastic Optimization: A Bayesian Perspective

We develop a principled approach to end-to-end learning in stochastic optimization. First, we show that the standard end-to-end learning algorithm admits a Bayesian interpretation and trains a posterior Bayes action map. Building on the insights of this analysis, we then propose new end-to-end learning algorithms for training decision maps that output solutions of empirical risk … Read more

Adaptive Importance Sampling Based Surrogation Methods for Bayesian Hierarchical Models, via Logarithmic Integral Optimization

We explore Maximum a Posteriori inference of Bayesian Hierarchical Models (BHMs) with intractable normalizers, which are increasingly prevalent in contemporary applications and pose computational challenges when combined with nonconvexity and nondifferentiability. To address these, we propose the Adaptive Importance Sampling-based Surrogation method, which efficiently handles nonconvexity and nondifferentiability while improving the sampling approximation of the … Read more

Sampling-based Decomposition Algorithms for Multistage Stochastic Programming

Sampling-based algorithms provide a practical approach to solving large-scale multistage stochastic programs. This chapter presents two alternative approaches to incorporating sampling within multistage stochastic linear programming algorithms. In the first approach, sampling is used to construct a sample average approximation (SAA) of the true multistage program. Subsequently, an optimization step is undertaken using deterministic decomposition … Read more

Sample average approximation and model predictive control for inventory optimization

We study multistage stochastic optimization problems using sample average approximation (SAA) and model predictive control (MPC) as solution approaches. MPC is frequently employed when the size of the problem renders stochastic dynamic programming intractable, but it is unclear how this choice affects out-of-sample performance. To compare SAA and MPC out-of-sample, we formulate and solve an … Read more

On the Regulatory and Economic Incentives for Renewable Hybrid Power Plants in Brazil

The complementarity between renewable generation profiles has been widely explored in the literature. Notwithstanding, the regulatory and economic frameworks for hybrid power plants add interesting challenges and opportunities for investors, regulators, and planners. Focusing on the Brazilian power market, this paper proposes a unified and isonomic firm energy certificate (FEC) calculation for non-controllable renewable generators, … Read more

Stochastic Ratios Tracking Algorithm for Large Scale Machine Learning Problems

Many machine learning applications and tasks rely on the stochastic gradient descent (SGD) algorithm and its variants. Effective step length selection is crucial for the success of these algorithms, which has motivated the development of algorithms such as ADAM or AdaGrad. In this paper, we propose a novel algorithm for adaptive step length selection in … Read more

Water resources management: A bibliometric analysis and future research directions

The stochastic dual dynamic programming (SDDP) algorithm introduced by Pereira and Pinto in 1991 has sparked essential research in the context of water resources management, mainly due to its ability to address large-scale multistage stochastic problems. This paper aims to provide a tutorial-type review of 32 years of research since the publication of the SDDP … Read more

Fair stochastic vehicle routing with partial deliveries

A common assumption in the models for the vehicle routing problem with stochastic demands is that all demands must be satisfied. This is achieved by including recourse actions in two-stage stochastic programming formulations or by ensuring with a high probability that all demand fits within the vehicle capacity (chance-constrained formulations). In this work, we relax … Read more

Differential Privacy via Distributionally Robust Optimization

In recent years, differential privacy has emerged as the de facto standard for sharing statistics of datasets while limiting the disclosure of private information about the involved individuals. This is achieved by randomly perturbing the statistics to be published, which in turn leads to a privacy-accuracy trade-off: larger perturbations provide stronger privacy guarantees, but they … Read more

Multi-model Partially Observable Markov Decision Processes

We propose a new multi-model partially observable Markov decision process (MPOMDP) model to address the issue of model ambiguity in partially observable Markov decision process. Here, model ambiguity is defined as the case where there are multiple credible optimization models with the same structure but different model parameters. The proposed MPOMDP model aims to learn … Read more