David L. Woodruff

In “Projective Hedging Algorithms for Multistage Stochastic Programming, Supporting Distributed and Asynchronous Implementation,” Eckstein, Watson, and Woodruff derive a new class of decomposition methods for convex multistage stochastic programs defined on finite but potentially large scenario trees. These methods resemble Rockafellar and Wets’ now-classical progressive hedging (PH) method but are based on a flexible projective operator-splitting scheme instead of the standard alternating direction method of multipliers (ADMM). The new algorithms only need to reoptimize subproblems for a subset of the scenarios at each iteration, instead of all of them, and are also amenable to a form of asynchronous implementation, without the algorithm randomization or small step-size requirements usually imposed in such contexts. In the online appendix, the authors demonstrate significant computational gains over PH, applying hundreds or thousands of processor cores to problem instances with up to a million scenarios.

In the context of optimization under uncertainty, we consider various combinations of distribution estimation and resampling (bootstrap and bagging) for obtaining samples used to estimate a confidence interval for an optimality gap. This paper makes three experimental contributions to on-going research in data driven stochastic programming: a) most of the combinations of distribution estimation and resampling result in algorithms that have not been published before, b) within the algorithms, we describe innovations that improve performance, and c) we provide open-source software implementations of the algorithms. Among others, three important conclusions can be drawn: using a smoothed point estimate for the optimality gap for the center of the confidence interval is preferable to a purely empirical estimate, bagging generally performs better than bootstrap, and smoothed bagging sometimes performs better than bagging based directly on the data.

Data-based, two-stage stochastic programming using bootstrap and bagging for confidence intervals.

Catastrophe-related insurance (e.g. business interruption insurance) is an effective financing tool for global corporations to reduce economic losses caused by high impact events. Flexible operational planning is an often-used tool enabling rapid adjustment of operational plans for reducing catastrophe-related damage costs. The interaction between catastrophe insurance and flexible operations planning has rarely been studied. In this paper, we develop a stochastic programming model for a multi-echelon global supply chain network that we solve to investigate the impact of purchasing catastrophe insurance on supply chain operational planning in a catastrophe-prone environment. Computational simulations are developed for evaluating solution quality and measuring catastrophe-related damage costs. We find that it may be optimal for supply chains to scrap redundant products in catastrophes when customer demand falls below the expected level. Purchasing catastrophe insurance may encourage supply chains to scrap more products, which results in more catastrophe-related damages. From analysing supply chain costs, catastrophe-related damage costs, and operational plans, we find that a higher compensation rate of catastrophe insurance triggers more production activities being planned at the vulnerable node just before the vulnerable time period, especially for low residual value products. Finally, we give managerial insights to help reduce unnecessary damages in catastrophes.

Cell2Fire is a new cell-based wildland fire growth simulator designed to integrate data-driven landscape management planning models. The fire environment is modeled by partitioning the landscape into cells characterized by fuel, weather, moisture content, and topographic attributes. The model can use existing fire spread models such as the Canadian Forest Fire Behavior Prediction System to model fire growth. Cell2Fire is structured to facilitate its use for predicting the growth of individual fires or by embedding it in landscape management simulation models. Decision-making models such as fuel treatment/harvesting plans can be easily integrated and evaluated. It incorporates a series of out-of-the-box planning heuristics that provide benchmarks for comparison. We illustrate their use by applying and evaluating a series of harvesting plans for forest landscapes in Canada. We validated Cell2Fire by using it to predict the growth of both real and hypothetical fires, comparing our predictions with the fire scars produced by a validated fire growth simulator (Prometheus). Cell2Fire is implemented as an open-source project that exploits parallelism to efficiently support the modeling of fire growth across large spatial and temporal scales. Our experiments indicate that Cell2Fire is able to efficiently simulate wildfires (up to 30x faster) under different conditions with similar accuracy as state-of-the-art simulators (above 90% of accuracy). We demonstrate its effectiveness as part of a harvest planning optimization framework, identifying relevant metrics to capture and actions to mitigate the impact of wildfire uncertainty.