The stochastic contextual bandit problem, recognized for its effectiveness in navigating the classic exploration-exploitation dilemma through ongoing player-environment interactions, has found broad applications across various industries. This utility largely stems from the algorithms’ ability to accurately forecast reward functions and maintain an optimal balance between exploration and exploitation, contingent upon the precise selection and calibration of hyperparameters. However, the inherently dynamic and real-time nature of bandit environments significantly complicates hyperparameter tuning, rendering traditional offline methods inadequate. While specialized methods have been developed to overcome these challenges, they often face three primary issues: difficulty in adaptively learning hyperparameters in ever-changing environments, inability to simultaneously optimize multiple hyperparameters for complex models, and inefficiencies in data utilization and knowledge transfer from analogous tasks. To tackle these hurdles, this paper introduces an innovative transfer learning-based approach designed to harness past task knowledge for accelerated optimization and dynamically optimize multiple hyperparameters, making it well-suited for fluctuating environments. The method employs a dual Gaussian meta-model strategy—one for transfer learning and the other for assessing hyperparameters’ performance within the current task —enabling it to leverage insights from previous tasks while quickly adapting to new environmental changes. Furthermore, the framework’s meta-model-centric architecture enables simultaneous optimization of multiple hyperparameters. Experimental evaluations demonstrate that this approach markedly outperforms competing methods in scenarios with perturbations and exhibits superior performance in 70% of stationary cases while matching performance in the remaining 30%. This superiority in performance, coupled with its computational efficiency on par with existing alternatives, positions it as a superior and practical solution for optimizing hyperparameters in contextual bandit settings.

Nowadays, online social networks play a crucial role in shaping human communication in various life activities. Social Network Analysis (SNA) provides valuable insights for businesses, authorities, and platform owners. One of the challenging tasks in SNA is detecting sequential change points in observed events in social networks when the parameters of statistical distribution of post-change networks are unknown. This challenging problem is particularly prominent in various real-world network systems, especially when the events in the networks can be modeled through a Hawkes process. Identifying change points in the stream of social network data, where the underlying statistical properties undergo significant changes, necessitates the development of adaptive online algorithms. Additionally, in cases where the use of maximum likelihood estimators is impractical or when no exact recursive function for likelihood is available, addressing this issue becomes more complex. This paper proposes likelihood estimators using online convex optimization methods, incorporating the adaptive moment estimation (ADAM) algorithm. The proposed method is seamlessly integrated into the sequential anomaly detection procedure for events in social networks. Experimental results on monitoring time between events demonstrate lower Expected Delay Detection (EDD), indicating the superiority of the proposed algorithm in both synthetic and real-world datasets such as Facebook and contact networks of individuals causing disease transmission. The proposed robust solution provides an efficient practical tool in situations where traditional methods face limitations in swift detection with high accuracy.

There have been many applications for machine learning algorithms in different fields. The importance of hyperparameters for machine learning algorithms is their control over the behaviors of training algorithms and their crucial impact on the performance of machine learning models. Tuning hyperparameters crucially affects the performance of machine learning algorithms, and future advances in this area mainly depend on well-tuned hyperparameters. Nevertheless, the high computational cost involved in evaluating the algorithms in large datasets or complicated models is a significant limitation that causes inefficiency of the tuning process. Besides, increased online applications of machine learning approaches have led to the requirement of producing good answers in less time. The present study first presents a novel classification of hyperparameter types based on their types to create high-quality solutions quickly. Then, based on this classification and using the hypergradient technique, some hyperparameters of deep learning algorithms are adjusted during the training process to decrease the search space and discover the optimal values of the hyperparameters. This method just needs only the parameters of the previous two steps and the gradient of the previous step. Finally, the proposed method is combined with other techniques in hyperparameter optimization, and the results are reviewed in two case studies. As confirmed by experimental results, the performance of the algorithms with the proposed method have been increased 36.62% and 23.16% (based on the best average accuracy) for Cifar10 and Cifar100 dataset respectively in early stages while the final produced answers with this method are equal to or better than the algorithms without it. Therefore, this method can be combined with hyperparameter optimization algorithms in order to improve their performance and make them more appropriate for online use by just using the parameters of the previous two steps and the gradient of the previous step.

In practical scheduling problems, some factors such as depreciation cost, green costs like the amount of energy consumption or carbon emission, other resources consumption, raw material cost, etc., are not explicitly related to the machine processing times. Most of these factors can be generally considered as machine costs. Considering the machine cost as another objective alongside the other classical time-driven decision objectives can be an attractive work in scheduling problems. However, this subject has not been discussed thoroughly in the literature for the case the machines have fixed processing costs. This paper investigates a general unrelated parallel machine scheduling problem with the machine processing cost. In this problem, it is assumed that processing a job on a machine incurs a particular cost in addition to processing time. The considered objectives are the makespan and the total cost, which are minimized simultaneously to obtain Pareto optimal solutions. The efficacy of the mathematical programming approach to solve the considered problem is evaluated rigorously in this paper. In this respect, a multiobjective solution procedure is proposed to generate a set of appropriate Pareto solutions for the decision-maker based on the mathematical programming approach. In this procedure, the ϵ-constraint method is first used to convert the bi-objective optimization problem into single-objective problems by transferring the makespan to the set of constraints. Then, the single-objective problems are solved using the CPLEX software. Moreover, some strategies are also used to reduce the solution time of the problem. At the end of the paper, comprehensive numerical experiments are conducted to evaluate the performance of the proposed multiobjective solution procedure. A vast range of problem sizes is selected for the test problems, up to 50 machines and 500 jobs. Furthermore, some rigorous analyses are performed to significantly restrict the patterns of generating processing time and cost parameters for the problem instances. The experimental results demonstrate the mathematical programming solution approach's efficacy in solving the problem. It is observed that even for large-scale problems, a diverse set of uniformly distributed Pareto solutions can be generated in a reasonable time with the gaps from the optimality less than 0.03 most of the time.

In this research, a maximal covering location problem (MCLP) with real-world constraints such as multiple types of facilities and vehicles with different setup costs is taken into account. An original mixed integer linear programming (MILP) model is constructed in order to find the optimal solution. Since the problem at hand is shown to be NP-hard, a constructive heuristic method and a meta-heuristic approach based on genetic algorithm (GA) are developed to solve the problem. To find the most effective solution technique, a set of problems of different sizes is randomly generated and solved by the proposed solution methods. Computational results demonstrate that the heuristic method is capable of producing optimal or near-optimal solutions in a rational execution time.

Machine vision systems are among the novel tools proven to be useful in different applications, among which monitoring and controlling manufacturing processes is one of the most important ones. However, due to the complexity resulted from high-dimensional image data and their inherent correlations, the acquisition of traditional statistical process control tools seems inapplicable. To overcome the shortcomings of the traditional methods in this regard, a statistical model is proposed in this paper which utilizes the concepts of both the PCA-based T2 control chart and the classification methods to develop a tool capable of controlling an image-based process. By defining the warning zones, collected data taken from an image-based process are classified into more than the two classes related to in-control and out-of-control processes. This helps practitioners to define rules to make it easier to realize when the process is getting out of control. Through simulation, the accuracy performance and the speed of four different types of classifiers including linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), kth nearest neighbors (KNN), and support vector machine (SVM) are assessed in different scenarios, based on which the functionality of the proposed approach is evaluated in in-control and out-of-control conditions.