Over-parameterization of deep neural networks (DNNs) has shown high prediction accuracy for many applications. Although effective, the large number of parameters hinders its popularity on resource-limited devices and has an outsize environmental impact. Sparse training (using a fixed number of nonzero weights in each iteration) could significantly mitigate the training costs by reducing the model size. However, existing sparse training methods mainly use either random-based or greedy-based drop-and-grow strategies, resulting in local minimal and low accuracy. In this work, we consider the dynamic sparse training as a sparse connectivity search problem and design an exploitation and exploration acquisition function to escape from local optima and saddle points. We further design an acquisition function and provide the theoretical guarantees for the proposed method and clarify its convergence property. Experimental results show that sparse models (up to 98\% sparsity) obtained by our proposed method outperform the SOTA sparse training methods on a wide variety of deep learning tasks. On VGG-19 / CIFAR-100, ResNet-50 / CIFAR-10, ResNet-50 / CIFAR-100, our method has even higher accuracy than dense models. On ResNet-50 / ImageNet, the proposed method has up to 8.2\% accuracy improvement compared to SOTA sparse training methods.

Since analytical solutions are usually not available, numerical solutions on discretized space and time domains are considered for predictive modeling. Leveraging state-of-the-art computational fluid dynamics (CFD) approaches based on finite volume [3] or finite element [4, 5] methods, one could obtain high-fidelity solutions that can be suitable for downstream design optimization and control purposes. However, the cost of performing such simulations is significant, and becomes prohibitively high for complex problems arising from real-world applications. This limitation of traditional CFD approaches has inspired the development of data-driven projectionbased reduced-order modeling techniques. Such models are usually used in an offline-online manner. In the offline stage, an approximation of the governing flow dynamics in a low-order linear subspace is constructed based on available fluid flow data collected. This approximation reduces the complexity of the problem in the online stage, making it possible to acquire fast, accurate predictions. Popular methods in this category include proper orthogonal decomposition (POD) [6, 7], dynamic mode decomposition (DMD) [8], along with many variants (e.g., [9, 10, 11]). However, these methods encounter difficulty when applied to scenarios with high Reynolds numbers and convection-dominated problems, whereas one needs a significantly large number of linear subspaces to achieve a satisfactory approximation.

To meet order fulfillment targets, manufacturers seek to optimize production schedules. Machine learning can support this objective by predicting throughput times on production lines given order specifications. However, this is challenging when manufacturers produce customized products because customization often leads to changes in the probability distribution of operational data -- so-called distributional shifts. Distributional shifts can harm the performance of predictive models when deployed to future customer orders with new specifications. The literature provides limited advice on how such distributional shifts can be addressed in operations management. Here, we propose a data-driven approach based on adversarial learning and job shop scheduling, which allows us to account for distributional shifts in manufacturing settings with high degrees of product customization. We empirically validate our proposed approach using real-world data from a job shop production that supplies large metal components to an oil platform construction yard. Across an extensive series of numerical experiments, we find that our adversarial learning approach outperforms common baselines. Overall, this paper shows how production managers can improve their decision-making under distributional shifts.

This work investigates the Monte Carlo Tree Search (MCTS) method combined with dedicated heuristics for solving the Weighted Vertex Coloring Problem. In addition to the basic MCTS algorithm, we study several MCTS variants where the conventional random simulation is replaced by other simulation strategies including greedy and local search heuristics. We conduct experiments on well-known benchmark instances to assess these combined MCTS variants. We provide empirical evidence to shed light on the advantages and limits of each simulation strategy. This is an extension of the work of Grelier and al. presented at EvoCOP2022.

In this study, we investigate the effectiveness of using the PID (Proportional - Integral - Derivative) control loop factors for modifying response thresholds in a decentralized, non-communicating, threshold-based swarm. Each agent in our swarm has a set of four thresholds, each corresponding to a task the agent is capable of performing. The agent will act on a particular task if the stimulus is higher than its corresponding threshold. The ability to modify their thresholds allows the agents to specialize dynamically in response to task demands. Current approaches to dynamic thresholds typically use a learning and forgetting process to adjust thresholds. These methods are able to effectively specialize once, but can have difficulty re-specializing if the task demands change. Our approach, inspired by the PID control loop, alters the threshold values based on the current task demand value, the change in task demand, and the cumulative sum of previous task demands. We show that our PID-inspired method is scalable and outperforms fixed and current learning and forgetting response thresholds with non-changing, constant, and abrupt changes in task demand. This superior performance is due to the ability of our method to re-specialize repeatedly in response to changing task demands.

Physics simulations are a computational bottleneck in computer-aided design (CAD) optimization processes. Hence, in order to make accurate (computationally expensive) simulations feasible for use in design optimization, one requires either an optimization framework that is highly sample-efficient or fast data-driven proxies (surrogate models) for long running simulations. In this work, we leverage recent advances in optimization and artificial intelligence (AI) to address both of these potential solutions, in the context of designing an optimal unmanned underwater vehicle (UUV). We first investigate and compare the sample efficiency and convergence behavior of different optimization techniques with a standard computational fluid dynamics (CFD) solver in the optimization loop. We then develop a deep neural network (DNN) based surrogate model to approximate drag forces that would otherwise be computed via direct numerical simulation with the CFD solver. The surrogate model is in turn used in the optimization loop of the hull design. Our study finds that the Bayesian Optimization Lower Condition Bound (BO LCB) algorithm is the most sample-efficient optimization framework and has the best convergence behavior of those considered. Subsequently, we show that our DNN-based surrogate model predicts drag force on test data in tight agreement with CFD simulations, with a mean absolute percentage error (MAPE) of 1.85%. Combining these results, we demonstrate a two-orders-of-magnitude speedup (with comparable accuracy) for the design optimization process when the surrogate model is used. To our knowledge, this is the first study applying Bayesian optimization and DNN-based surrogate modeling to the problem of UUV design optimization, and we share our developments as open-source software.

Generalizability and stability are two key objectives for operating reinforcement learning (RL) agents in the real world. Designing RL algorithms that optimize these objectives can be a costly and painstaking process. This paper presents MetaPG, an evolutionary method for automated design of actor-critic loss functions. MetaPG explicitly optimizes for generalizability and performance, and implicitly optimizes the stability of both metrics. We initialize our loss function population with Soft Actor-Critic (SAC) and perform multi-objective optimization using fitness metrics encoding single-task performance, zero-shot generalizability to unseen environment configurations, and stability across independent runs with different random seeds. On a set of continuous control tasks from the Real-World RL Benchmark Suite, we find that our method, using a single environment during evolution, evolves algorithms that improve upon SAC's performance and generalizability by 4% and 20%, respectively, and reduce instability up to 67%. Then, we scale up to more complex environments from the Brax physics simulator and replicate generalizability tests encountered in practical settings, such as different friction coefficients. MetaPG evolves algorithms that can obtain 10% better generalizability without loss of performance within the same meta-training environment and obtain similar results to SAC when doing cross-domain evaluations in other Brax environments. The evolution results are interpretable; by analyzing the structure of the best algorithms we identify elements that help optimizing certain objectives, such as regularization terms for the critic loss.

We study the problem of optimally routing plug-in electric and conventional fuel vehicles on a city level. In our model, commuters selfishly aim to minimize a local cost that combines travel time, from a fixed origin to a desired destination, and the monetary cost of using city facilities, parking or service stations. The traffic authority can influence the commuters' preferred routing choice by means of personalized discounts on parking tickets and on the energy price at service stations. We formalize the problem of designing these monetary incentives optimally as a large-scale bilevel game, where constraints arise at both levels due to the finite capacities of city facilities and incentives budget. Then, we develop an efficient decentralized solution scheme with convergence guarantees based on BIG Hype, a recently-proposed hypergradient-based algorithm for hierarchical games. Finally, we validate our model via numerical simulations over the Anaheim's network, and show that the proposed approach produces sensible results in terms of traffic decongestion and it is able to solve in minutes problems with more than 48000 variables and 110000 constraints.

Transformer fault diagnosis (TFD) is a critical aspect of power system maintenance and management. This review paper provides a comprehensive overview of the current state of the art in TFD using artificial intelligence (AI) and dissolved gas analysis (DGA). The paper presents an analysis of recent advancements in this field, including the use of deep learning algorithms and advanced data analytics techniques, and their potential impact on TFD and the power industry as a whole. The review also highlights the benefits and limitations of different approaches to transformer fault diagnosis, including rule-based systems, expert systems, neural networks, and machine learning algorithms. Overall, this review aims to provide valuable insights into the importance of TFD and the role of AI in ensuring the reliable operation of power systems.

This paper investigates policy resilience to training-environment poisoning attacks on reinforcement learning (RL) policies, with the goal of recovering the deployment performance of a poisoned RL policy. Due to the fact that the policy resilience is an add-on concern to RL algorithms, it should be resource-efficient, time-conserving, and widely applicable without compromising the performance of RL algorithms. This paper proposes such a policy-resilience mechanism based on an idea of knowledge sharing. We summarize the policy resilience as three stages: preparation, diagnosis, recovery. Specifically, we design the mechanism as a federated architecture coupled with a meta-learning manner, pursuing an efficient extraction and sharing of the environment knowledge. With the shared knowledge, a poisoned agent can quickly identify the deployment condition and accordingly recover its policy performance. We empirically evaluate the resilience mechanism for both model-based and model-free RL algorithms, showing its effectiveness and efficiency in restoring the deployment performance of a poisoned policy.