Mostafa Shamsi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Mostafa Shamsi has a total of 63 co-authors. Here are the top co-authors by number of shared publications:
- Mehdi Dehghan
Department of Applied Mathematics, Faculty of Mathematics and Computer Sciences, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran
6 shared publications - Zahra Foroozandeh
Polytechnic of Porto
5 shared publications - Maria Do Rosário de Pinho
Faculdade de Engenharia da Universidade do Porto, DEEC, SYSTEC, Porto, Portugal
4 shared publications - Delfim F. Marado Torres
Center for Research and Development in Mathematics and Applications (CIDMA), Department of Mathematics, University of Aveiro, 3810-193 Aveiro, Portugal
3 shared publications Hassan Khosravian-Arab
Department of Applied Mathematics
3 shared publications- Shahrokh Esmaeili
Department of Applied Mathematics, University of Kurdistan, Sanandaj, Iran
3 shared publications
Latest Publications
- 30 July 2026
- Published by Elsevier in Neurocomputing
- Vol. 702
HBPP: A transformer-inspired architecture for physics-informed learning of parametric PDEs
- 30 July 2026
- Published by Elsevier in Neurocomputing
- Vol. 702
HBPP: A transformer-inspired architecture for physics-informed learning of parametric PDEs
- Open Access
- 10 February 2026
- Published by Elsevier in SSRN Electronic Journal
A Comprehensive Review of Meta-Learning: From Theoretical Foundations to Practical Applications
Meta-learning, often characterized as "learning to learn," addresses a fundamental limitation of conventional deep learning: the requirement for large amounts of labeled data to achieve good performance on new tasks. By training models on a distribution of related tasks rather than a single task, meta-learning enables rapid adaptation to novel tasks using only a handful of examples. This paradigm has gained significant attention due to its potential to reduce data requirements and build more flexible, adaptive AI systems. This survey provides a comprehensive treatment of modern neural meta-learning, emphasizing both theoretical foundations and practical applications. We formalize meta-learning as a bilevel optimization problem and present a detailed taxonomy of approaches, including metric-based methods that learn embedding spaces, model-based methods that internalize adaptation in network architectures, and optimization-based methods that learn initializations or update rules. We examine connections between meta-learning and related paradigms such as transfer learning, multi-task learning, and continual learning, clarifying the distinctive characteristics of each. We dedicate substantial coverage to applications across diverse domains including computer vision (few-shot classification, object detection, domain adaptation), reinforcement learning (meta-RL algorithms and robotics), natural language processing (few-shot text classification, low-resource translation, in-context learning), and specialized areas such as drug discovery, medical imaging, and federated learning. Beyond surveying existing methods, we provide critical analysis of current challenges including meta-overfitting, computational efficiency, and evaluation protocols. We discuss theoretical developments in understanding meta-generalization through PAC-Bayes bounds and information-theoretic analyses, and identify promising research directions for advancing toward more data-efficient and adaptable artificial intelligence systems.
- 10 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Comprehensive Review of Meta-Learning: From Theoretical Foundations to Practical Applications
Meta-learning, often characterized as "learning to learn," addresses a fundamental limitation of conventional deep learning: the requirement for large amounts of labeled data to achieve good performance on new tasks. By training models on a distribution of related tasks rather than a single task, meta-learning enables rapid adaptation to novel tasks using only a handful of examples. This paradigm has gained significant attention due to its potential to reduce data requirements and build more flexible, adaptive AI systems. This survey provides a comprehensive treatment of modern neural meta-learning, emphasizing both theoretical foundations and practical applications. We formalize meta-learning as a bilevel optimization problem and present a detailed taxonomy of approaches, including metric-based methods that learn embedding spaces, model-based methods that internalize adaptation in network architectures, and optimization-based methods that learn initializations or update rules. We examine connections between meta-learning and related paradigms such as transfer learning, multi-task learning, and continual learning, clarifying the distinctive characteristics of each. We dedicate substantial coverage to applications across diverse domains including computer vision (few-shot classification, object detection, domain adaptation), reinforcement learning (meta-RL algorithms and robotics), natural language processing (few-shot text classification, low-resource translation, in-context learning), and specialized areas such as drug discovery, medical imaging, and federated learning. Beyond surveying existing methods, we provide critical analysis of current challenges including meta-overfitting, computational efficiency, and evaluation protocols. We discuss theoretical developments in understanding meta-generalization through PAC-Bayes bounds and information-theoretic analyses, and identify promising research directions for advancing toward more data-efficient and adaptable artificial intelligence systems.
- 29 October 2025
- Published by World Scientific Pub Co Pte Ltd in Fluctuation and Noise Letters
- Vol. 25 (01)
Parameter Estimation for Tempered Stable Distributions Using Cumulant Matching
This paper presents a novel parameter estimation method for tempered stable distributions based on cumulant matching. Estimation techniques, such as Maximum Likelihood Estimation (MLE), face significant challenges due to the intractable density functions of tempered stable distributions. To address this issue, we introduce the Method of Cumulants (MoC), which leverages the analytically simpler cumulants of these distributions for parameter estimation. We derive explicit formulas for the cumulants of classical tempered stable distributions and establish the statistical consistency of the MoC estimator. Through extensive simulations, we demonstrate the efficiency and accuracy of our method, with results indicating low mean squared errors across various parameter settings. This approach not only enhances the theoretical framework for parameter estimation in tempered stable distributions but also shows promising practical applications, particularly in finance. Future work will explore its applicability to other types of tempered stable distributions.
- 29 October 2025
- Published by World Scientific Pub Co Pte Ltd in Fluctuation and Noise Letters
- Vol. 25 (01)
Parameter Estimation for Tempered Stable Distributions Using Cumulant Matching
This paper presents a novel parameter estimation method for tempered stable distributions based on cumulant matching. Estimation techniques, such as Maximum Likelihood Estimation (MLE), face significant challenges due to the intractable density functions of tempered stable distributions. To address this issue, we introduce the Method of Cumulants (MoC), which leverages the analytically simpler cumulants of these distributions for parameter estimation. We derive explicit formulas for the cumulants of classical tempered stable distributions and establish the statistical consistency of the MoC estimator. Through extensive simulations, we demonstrate the efficiency and accuracy of our method, with results indicating low mean squared errors across various parameter settings. This approach not only enhances the theoretical framework for parameter estimation in tempered stable distributions but also shows promising practical applications, particularly in finance. Future work will explore its applicability to other types of tempered stable distributions.
- Open Access
- 4 March 2025
- Published by MDPI AG in International Journal of Financial Studies
- Vol. 13 (1) , 38
Robust Portfolio Selection Under Model Ambiguity Using Deep Learning
In this study, we address the ambiguity in portfolio optimization, particularly focusing on the uncertainty related to the statistical parameters governing asset returns. We propose a novel method that combines robust optimization with artificial neural networks (ANNs). Our approach effectively handles both the randomness inherent in asset prices and the ambiguity in their governing parameters. Through our method, we consider both simulated data, using the Exponential Ornstein–Uhlenbeck process, and real-world stock price data. The results showcase that our ANN-based method outperforms traditional benchmark methods such as equally weighted portfolio and adaptive mean–variance portfolio selection.
- 4 March 2025
- Published by MDPI AG in International Journal of Financial Studies
- Vol. 13 (1) , 38
Open AccessRobust Portfolio Selection Under Model Ambiguity Using Deep Learning
In this study, we address the ambiguity in portfolio optimization, particularly focusing on the uncertainty related to the statistical parameters governing asset returns. We propose a novel method that combines robust optimization with artificial neural networks (ANNs). Our approach effectively handles both the randomness inherent in asset prices and the ambiguity in their governing parameters. Through our method, we consider both simulated data, using the Exponential Ornstein–Uhlenbeck process, and real-world stock price data. The results showcase that our ANN-based method outperforms traditional benchmark methods such as equally weighted portfolio and adaptive mean–variance portfolio selection.
- 12 March 2024
- Published by SAGE Publications in Journal of Vibration and Control
- Vol. 31 (7-8) , 1126-1140
A direct transcription method for solving distributed-order fractional optimal control problems
This paper considers the numerical solution of distributed-order Fractional Optimal Control Problems (FOCPs). The common approaches for solving these problems involve approximating the problem with a multi-term FOCP and subsequently discretizing the resulting multi-term problem using either a direct or indirect method. However, this paper bypasses the need for approximation with a multi-term FOCP by introducing innovative approximation formulas for distributed-order fractional derivatives. These approximation formulas are directly derived from the conventional Grünwald–Letnikov, L1, and trapezoidal formulas. Building upon these formulas, we develop the relevant fractional-order distributed-order operational matrices. Utilizing the operational matrix, we easily discretize the distributed-order FOCP to a Non-Linear Programming (NLP), which can be solved by a suitable NLP-solver. Additionally, to enhance the efficiency of solving the resulting NLP, we determine the closed-forms of both the Jacobian of constraints and the gradient of the objective function. The presented method is characterized by its speed, simplicity, and ease of implementation. Moreover, it can be used to solve a wide range of distributed-order FOCPs, such as those involving nonlinear dynamics, free final times, free terminal conditions, and path constraints. By means of ample numerical tests, the accuracy and efficiency of the proposed method are assessed.
- 12 March 2024
- Published by SAGE Publications in Journal of Vibration and Control
- Vol. 31 (7-8) , 1126-1140
A direct transcription method for solving distributed-order fractional optimal control problems
This paper considers the numerical solution of distributed-order Fractional Optimal Control Problems (FOCPs). The common approaches for solving these problems involve approximating the problem with a multi-term FOCP and subsequently discretizing the resulting multi-term problem using either a direct or indirect method. However, this paper bypasses the need for approximation with a multi-term FOCP by introducing innovative approximation formulas for distributed-order fractional derivatives. These approximation formulas are directly derived from the conventional Grünwald–Letnikov, L1, and trapezoidal formulas. Building upon these formulas, we develop the relevant fractional-order distributed-order operational matrices. Utilizing the operational matrix, we easily discretize the distributed-order FOCP to a Non-Linear Programming (NLP), which can be solved by a suitable NLP-solver. Additionally, to enhance the efficiency of solving the resulting NLP, we determine the closed-forms of both the Jacobian of constraints and the gradient of the objective function. The presented method is characterized by its speed, simplicity, and ease of implementation. Moreover, it can be used to solve a wide range of distributed-order FOCPs, such as those involving nonlinear dynamics, free final times, free terminal conditions, and path constraints. By means of ample numerical tests, the accuracy and efficiency of the proposed method are assessed.
- 14 February 2022
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 37 (4) , 1415-1446
A quasi-Newton proximal bundle method using gradient sampling technique for minimizing nonsmooth convex functions
This study aims to merge the well-established ideas of bundle and Gradient Sampling (GS) methods to develop an algorithm for locating a minimizer of a nonsmooth convex function. In the proposed method, with the help of the GS technique, we sample a number of differentiable auxiliary points around the current iterate. Then, by applying the standard techniques used in bundle methods, we construct a polyhedral (piecewise linear) model of the objective function. Moreover, by performing quasi-Newton updates on the set of auxiliary points, this polyhedral model is augmented with a regularization term that enjoys second-order information. If required, this initial model is improved by the techniques frequently used in GS and bundle methods. We analyse the global convergence of the proposed method. As opposed to the original GS method and some of its variants, our convergence analysis is independent of the size of the sample. In our numerical experiments, various aspects of the proposed method are examined using a variety of test problems. In particular, in contrast with many variants of bundle methods, we will see that the user can supply gradients approximately. Moreover, we compare the proposed method with some efficient variants of GS and bundle methods.
- 14 February 2022
- Published by Taylor & Francis in Optimization Methods and Software
- Vol. 37 (4) , 1415-1446
A quasi-Newton proximal bundle method using gradient sampling technique for minimizing nonsmooth convex functions
This study aims to merge the well-established ideas of bundle and Gradient Sampling (GS) methods to develop an algorithm for locating a minimizer of a nonsmooth convex function. In the proposed method, with the help of the GS technique, we sample a number of differentiable auxiliary points around the current iterate. Then, by applying the standard techniques used in bundle methods, we construct a polyhedral (piecewise linear) model of the objective function. Moreover, by performing quasi-Newton updates on the set of auxiliary points, this polyhedral model is augmented with a regularization term that enjoys second-order information. If required, this initial model is improved by the techniques frequently used in GS and bundle methods. We analyse the global convergence of the proposed method. As opposed to the original GS method and some of its variants, our convergence analysis is independent of the size of the sample. In our numerical experiments, various aspects of the proposed method are examined using a variety of test problems. In particular, in contrast with many variants of bundle methods, we will see that the user can supply gradients approximately. Moreover, we compare the proposed method with some efficient variants of GS and bundle methods.
- 31 August 2020
- Published by Springer Nature in Journal of Optimization Theory and Applications
- Vol. 187 (1) , 181-204
A Gradient Sampling Method Based on Ideal Direction for Solving Nonsmooth Optimization Problems
- 31 August 2020
- Published by Springer Nature in Journal of Optimization Theory and Applications
- Vol. 187 (1) , 181-204
A Gradient Sampling Method Based on Ideal Direction for Solving Nonsmooth Optimization Problems
- 20 July 2020
- Published by Springer Nature in Computational Optimization and Applications
- Vol. 77 (2) , 379-409
A new method based on the proximal bundle idea and gradient sampling technique for minimizing nonsmooth convex functions
- 20 July 2020
- Published by Springer Nature in Computational Optimization and Applications
- Vol. 77 (2) , 379-409
A new method based on the proximal bundle idea and gradient sampling technique for minimizing nonsmooth convex functions