Contextual decision-making problems have witnessed extensive applications in various fields such as online content recommendation, personalized healthcare, and autonomous vehicles, where a core practical challenge is to select a suitable surrogate model for capturing unknown complicated reward functions. It is often the case that both high approximation accuracy and explicit uncertainty quantification are desired. In this work, we propose a neural network-accompanied Gaussian process (NN-AGP) model, which leverages neural networks to approximate the unknown and potentially complicated reward function regarding the contextual variable, and maintains a Gaussian process surrogate model with respect to the decision variable. Our model is shown to outperform existing approaches by offering better approximation accuracy thanks to the use of neural networks and possessing explicit uncertainty quantification from the Gaussian process. We also analyze the maximum information gain of the NN-AGP model and prove regret bounds for the corresponding algorithms. Moreover, we conduct experiments on both synthetic and practical problems, illustrating the effectiveness of our approach.

We consider three parallel service models in which customers of several types are served by several types of servers subject to a bipartite compatibility graph, and the service policy is first come first served. Two of the models have a fixed set of servers. The first is a queueing model in which arriving customers are assigned to the longest idling compatible server if available, or else queue up in a single queue, and servers that become available pick the longest waiting compatible customer, as studied by Adan and Weiss, 2014. The second is a redundancy service model where arriving customers split into copies that queue up at all the compatible servers, and are served in each queue on FCFS basis, and leave the system when the first copy completes service, as studied by Gardner et al., 2016. The third model is a matching queueing model with a random stream of arriving servers. Arriving customers queue in a single queue and arriving servers match with the first compatible customer and leave immediately with the customer, or they leave without a customer. The last model is relevant to organ transplants, to housing assignments, to adoptions and many other situations. We study the relations between these models, and show that they are closely related to the FCFS infinite bipartite matching model, in which two infinite sequences of customers and servers of several types are matched FCFS according to a bipartite compatibility graph, as studied by Adan et al., 2017. We also introduce a directed bipartite matching model in which we embed the queueing systems. This leads to a generalization of Burke’s theorem to parallel service systems.

We study the M/D/1 queue when jobs have firm deadlines for waiting (or sojourn) time. If a deadline is not met, a job-specific deadline violation cost is incurred. We derive explicit value functions for this M/D/1 queue that enable the development of efficient cost-aware dispatching policies to parallel servers. The performance of the resulting dispatching policies is evaluated by means of simulations.

In many service systems with multiple types of customers, providing server flexibility, e.g., by cross-training servers, is very expensive. On the other hand, there is often inherent flexibility in some of the customers that is not used by the system. I argue that taking advantage of such flexibility can create a win-win-win situation, in which overall performance can be greatly improved, and in which both flexible and non-flexible customers benefit. Moreover, only a small subset of customers needs to be flexible to obtain nearly the benefit of full flexibility.

An overview of distributed simulation of discrete event systems and the issues associated with it. Alternative approaches for decomposing the simulation into tasks that can be run on separate processors are described, and the potential parallelism associated with certain kinds of decomposition is studied. Existing synchronization algorithms are discussed. An attempt is made throughout to show what decomposition approaches and synchronization algorithms may be appropriate depending on properties of the application and the multiprocessor architecture. Empirical and analytical performance studies are described, where available.>