Luigi Rovito
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Luigi Rovito has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Andrea De Lorenzo
Department of Engineering and Architecture, University of Trieste, 34127 Trieste, Italy
11 shared publications - Luca Manzoni
Department of Mathematics, Informatics and Geosciences, University of Trieste, Trieste, Italy
11 shared publications - Damiano Ravalico
University of Trieste, trieste, Italy
2 shared publications - Davide Farinati
Nova Information Management School (Nova IMS), Universidade Nova de Lisboa, Lisboa, Portugal
2 shared publications - Eric Medvet
University of Trieste
2 shared publications - Giorgia Nadizar
University of Toulouse Capitole, Toulouse, France
2 shared publications
Latest Publications
- Open Access
- 1 September 2026
- Published by Walter de Gruyter GmbH in Cybernetics and Information Technologies
- Vol. 26 (3) , 229-246
Automated Digital Filter Synthesis with Canonical Signed Digit via Genetic Algorithms
Finite Impulse Response (FIR) digital filters are widely used in signal processing to remove noise and extract frequency-domain features. Genetic Algorithms (GAs) have been employed to synthesize FIR filters by optimizing the coefficients in their transfer function. A notable approach is to evolve FIR filter coefficients in the Canonical Signed Digit (CSD) representation, which simplifies hardware implementation by replacing multipliers with shifts and additions. However, its crossover and mutation operators frequently produce offspring that violate CSD constraints, requiring repeated retries or coefficient resets. We propose modified crossover and mutation operators that guarantee CSD-compliant offspring by construction. We compare our method against the original CSD-based GA and a real-valued GA baseline across five low-pass filter design problems, varying filter order and coefficient word length. Results show that our method consistently outperforms the original CSD-based approach.
- 1 September 2026
- Published by Walter de Gruyter GmbH in Cybernetics and Information Technologies
- Vol. 26 (3) , 229-246
Open AccessAutomated Digital Filter Synthesis with Canonical Signed Digit via Genetic Algorithms
Finite Impulse Response (FIR) digital filters are widely used in signal processing to remove noise and extract frequency-domain features. Genetic Algorithms (GAs) have been employed to synthesize FIR filters by optimizing the coefficients in their transfer function. A notable approach is to evolve FIR filter coefficients in the Canonical Signed Digit (CSD) representation, which simplifies hardware implementation by replacing multipliers with shifts and additions. However, its crossover and mutation operators frequently produce offspring that violate CSD constraints, requiring repeated retries or coefficient resets. We propose modified crossover and mutation operators that guarantee CSD-compliant offspring by construction. We compare our method against the original CSD-based GA and a real-valued GA baseline across five low-pass filter design problems, varying filter order and coefficient word length. Results show that our method consistently outperforms the original CSD-based approach.
- Open Access
- 2 April 2026
- Published by Elsevier in Applied Computing and Geosciences
- Vol. 30
An interpretable machine learning approach for alkalinity reconstruction in the Mediterranean Sea
- 2 April 2026
- Published by Elsevier in Applied Computing and Geosciences
- Vol. 30
Open AccessAn interpretable machine learning approach for alkalinity reconstruction in the Mediterranean Sea
- Open Access
- 17 January 2026
- Published by Scientific Oasis in Decision Making Advances
- Vol. 4 (1) , 337-359
Evolving Balanced Non-Linear Boolean Functions with Cellular Genetic Algorithms
Boolean functions are essential building blocks in the design of secure stream ciphers. Finding cryptographically strong Boolean functions is, however, a notoriously difficult optimization problem, owing to the size of the search space and the combinatorial nature of the task. Evolutionary Algorithms (EAs) have proven effective at constructing Boolean functions that are simultaneously balanced and highly nonlinear, two properties that are crucial for resisting correlation and statistical attacks. Nevertheless, these and other heuristic approaches often converge prematurely to local optima because of the complexity of the fitness landscape. Cellular Automata (CA) have been used extensively to represent and analyze Boolean functions, and spatially structured populations arranged on cellular toroidal grids have been shown, in other evolutionary paradigms, to slow the propagation of dominant individuals and to encourage broader exploration. The two ideas have not been brought together: cellular population structures have not previously been used to guide the search for balanced, nonlinear Boolean functions. We therefore investigate the integration of cellular population structures into Genetic Algorithms (GAs) for evolving balanced truth tables with high nonlinearity. We evolve Boolean functions of 8 to 16 variables and compare different neighborhood radii of the cellular grid. Our results show that, for larger numbers of variables, small neighborhoods improve the nonlinearity of the evolved functions relative to a non-spatial GA, and that this improvement is statistically significant. We emphasize that the contribution of this work is an analysis of the search and diversity dynamics induced by the cellular structure: the nonlinearity values obtained remain below the best-known values for balanced Boolean functions, and the study is not intended as a step forward in the construction of cryptographically deployable functions.
- 17 January 2026
- Published by Scientific Oasis in Decision Making Advances
- Vol. 4 (1) , 337-359
Open AccessEvolving Balanced Non-Linear Boolean Functions with Cellular Genetic Algorithms
Boolean functions are essential building blocks in the design of secure stream ciphers. Finding cryptographically strong Boolean functions is, however, a notoriously difficult optimization problem, owing to the size of the search space and the combinatorial nature of the task. Evolutionary Algorithms (EAs) have proven effective at constructing Boolean functions that are simultaneously balanced and highly nonlinear, two properties that are crucial for resisting correlation and statistical attacks. Nevertheless, these and other heuristic approaches often converge prematurely to local optima because of the complexity of the fitness landscape. Cellular Automata (CA) have been used extensively to represent and analyze Boolean functions, and spatially structured populations arranged on cellular toroidal grids have been shown, in other evolutionary paradigms, to slow the propagation of dominant individuals and to encourage broader exploration. The two ideas have not been brought together: cellular population structures have not previously been used to guide the search for balanced, nonlinear Boolean functions. We therefore investigate the integration of cellular population structures into Genetic Algorithms (GAs) for evolving balanced truth tables with high nonlinearity. We evolve Boolean functions of 8 to 16 variables and compare different neighborhood radii of the cellular grid. Our results show that, for larger numbers of variables, small neighborhoods improve the nonlinearity of the evolved functions relative to a non-spatial GA, and that this improvement is statistically significant. We emphasize that the contribution of this work is an analysis of the search and diversity dynamics induced by the cellular structure: the nonlinearity values obtained remain below the best-known values for balanced Boolean functions, and the study is not intended as a step forward in the construction of cryptographically deployable functions.
- Open Access
- 19 November 2025
- Published by Springer Nature in Genetic Programming and Evolvable Machines
- Vol. 26 (2) , 27
Semantic-based recombination and mutation in cellular-inspired genetic programming
The incorporation of a Cellular Automata (CA)-like structure into the population of Evolutionary Algorithms (EAs) has been shown to enhance solution quality. However, research on CA-like structures in the context of Genetic Programming (GP) remains limited. This work examines the impact of introducing such structures in Geometric Semantic variants of GP, specifically focusing on the well-established Geometric Semantic GP (GSGP) and the recently proposed SLIM-GSGP, which prioritizes generating smaller and more interpretable individuals. Furthermore, we analyze how cellular structures influence the effectiveness of semantic-based recombination and mutation in both GSGP and SLIM-GSGP. To this end, we conduct a comprehensive evaluation of these genetic operators, examining their effects both individually and in combination. We provide insights into how CA-like structures and semantic genetic operators influence both the quality and size of solutions in GSGP and SLIM-GSGP, offering a clear understanding of the trade-offs associated with these approaches.
- 19 November 2025
- Published by Springer Nature in Genetic Programming and Evolvable Machines
- Vol. 26 (2) , 27
Open AccessSemantic-based recombination and mutation in cellular-inspired genetic programming
The incorporation of a Cellular Automata (CA)-like structure into the population of Evolutionary Algorithms (EAs) has been shown to enhance solution quality. However, research on CA-like structures in the context of Genetic Programming (GP) remains limited. This work examines the impact of introducing such structures in Geometric Semantic variants of GP, specifically focusing on the well-established Geometric Semantic GP (GSGP) and the recently proposed SLIM-GSGP, which prioritizes generating smaller and more interpretable individuals. Furthermore, we analyze how cellular structures influence the effectiveness of semantic-based recombination and mutation in both GSGP and SLIM-GSGP. To this end, we conduct a comprehensive evaluation of these genetic operators, examining their effects both individually and in combination. We provide insights into how CA-like structures and semantic genetic operators influence both the quality and size of solutions in GSGP and SLIM-GSGP, offering a clear understanding of the trade-offs associated with these approaches.
- Open Access
- 22 August 2025
- Published by Springer Nature in SN Computer Science
- Vol. 6 (7) , 1-21
Exploring the Effect of Genetic Improvement for Large Language Models-Generated Code
In recent years, the field of Natural Language Processing (NLP) has made considerable progress with the development of neural network-based models, leading to the creation of various Large Language Models (LLMs). These models have demonstrated strong performance in various NLP tasks, such as language translation, sentiment analysis, and named entity recognition. One notable application of LLMs is their ability to generate code automatically from simple problem descriptions. However, even advanced LLMs frequently generate incorrect code. To address this issue, we extend a recently proposed method that aims to improve the correctness of code generated by LLMs using an evolutionary approach known as Genetic Improvement (GI). Our method involves constructing a dynamic grammar based on the LLM-generated code and using a problem-agnostic fitness function. In our experiments, we evaluated the proposed method on 25 well-known and widely-used problems across four different LLMs, both open-source and proprietary models. We demonstrate that our approach significantly improves the accuracy of code generated by LLMs. Specifically, for problems that the LLM alone does not fully solve, we show that GI significantly improves the initial LLM-generated solution in 50% to 75% of cases across the tested models. Our proposed GI approach remains effective as long as the initial LLM-generated code, despite some errors, provides a solid foundation for constructing a correct program.
- 22 August 2025
- Published by Springer Nature in SN Computer Science
- Vol. 6 (7) , 1-21
Open AccessExploring the Effect of Genetic Improvement for Large Language Models-Generated Code
In recent years, the field of Natural Language Processing (NLP) has made considerable progress with the development of neural network-based models, leading to the creation of various Large Language Models (LLMs). These models have demonstrated strong performance in various NLP tasks, such as language translation, sentiment analysis, and named entity recognition. One notable application of LLMs is their ability to generate code automatically from simple problem descriptions. However, even advanced LLMs frequently generate incorrect code. To address this issue, we extend a recently proposed method that aims to improve the correctness of code generated by LLMs using an evolutionary approach known as Genetic Improvement (GI). Our method involves constructing a dynamic grammar based on the LLM-generated code and using a problem-agnostic fitness function. In our experiments, we evaluated the proposed method on 25 well-known and widely-used problems across four different LLMs, both open-source and proprietary models. We demonstrate that our approach significantly improves the accuracy of code generated by LLMs. Specifically, for problems that the LLM alone does not fully solve, we show that GI significantly improves the initial LLM-generated solution in 50% to 75% of cases across the tested models. Our proposed GI approach remains effective as long as the initial LLM-generated code, despite some errors, provides a solid foundation for constructing a correct program.
- Open Access
- 13 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 453-462
Interpretable Non-linear Survival Analysis with Evolutionary Symbolic Regression
- 13 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 453-462
Open AccessInterpretable Non-linear Survival Analysis with Evolutionary Symbolic Regression
- 18 April 2025
- Published by Springer Nature
- p. 120-138
Exploring the Integration of Cellular Structures in Genetic Programming-Based Methods
- 18 April 2025
- Published by Springer Nature
- p. 120-138
Exploring the Integration of Cellular Structures in Genetic Programming-Based Methods
- Open Access
- 14 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 11-12
Interpretable Control Competition
Safety-critical applications rely on control systems-managed components, which not only have demanding requirements in terms of accuracy, but also require transparency and interpretability, to ensure trustworthiness. However, interpretability is often overlooked in favor of performance as (i) people perceive it as secondary, and (ii) there are no objective methods for measuring it. To address these shortcomings and raise awareness about the importance of interpretability in control systems, we propose the Interpretable Control Competition at GECCO 2024. We invite participants to solve two control problems, a discrete and a continuous one, with interpretable policies. We assess the policies in terms of both performance and interpretability, where the latter is evaluated by a pool of domain experts. Our overarching goal is to collect the contributions in a report, which should provide a guideline for practitioners in the field, gathering an array of state-of-the-art methods for interpretable control, together with suitable metrics for assessing the interpretability of a policy.
- 14 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 11-12
Open AccessInterpretable Control Competition
Safety-critical applications rely on control systems-managed components, which not only have demanding requirements in terms of accuracy, but also require transparency and interpretability, to ensure trustworthiness. However, interpretability is often overlooked in favor of performance as (i) people perceive it as secondary, and (ii) there are no objective methods for measuring it. To address these shortcomings and raise awareness about the importance of interpretability in control systems, we propose the Interpretable Control Competition at GECCO 2024. We invite participants to solve two control problems, a discrete and a continuous one, with interpretable policies. We assess the policies in terms of both performance and interpretability, where the latter is evaluated by a pool of domain experts. Our overarching goal is to collect the contributions in a report, which should provide a guideline for practitioners in the field, gathering an array of state-of-the-art methods for interpretable control, together with suitable metrics for assessing the interpretability of a policy.
- 28 March 2024
- Published by Springer Nature
- p. 108-124
Enhancing Large Language Models-Based Code Generation by Leveraging Genetic Improvement
- 28 March 2024
- Published by Springer Nature
- p. 108-124
Enhancing Large Language Models-Based Code Generation by Leveraging Genetic Improvement
- Open Access
- 23 February 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Evolutionary Learning and Optimization
- Vol. 4 (1) , 1-30
An Analysis of the Ingredients for Learning Interpretable Symbolic Regression Models with Human-in-the-loop and Genetic Programming
Interpretability is a critical aspect to ensure a fair and responsible use of machine learning (ML) in high-stakes applications. Genetic programming (GP) has been used to obtain interpretable ML models because it operates at the level of functional building blocks: if these building blocks are interpretable, there is a chance that their composition (i.e., the entire ML model) is also interpretable. However, the degree to which a model is interpretable depends on the observer. Motivated by this, we study a recently-introduced human-in-the-loop system that allows the user to steer GP’s generation process to their preferences, which shall be online-learned by an artificial neural network (ANN). We focus on the generation of ML models as analytical functions (i.e., symbolic regression) as this is a key problem in interpretable ML, and propose a two-fold contribution. First, we devise more general representations for the ML models for the ANN to learn upon, to enable the application of the system to a wider range of problems. Second, we delve into a deeper analysis of the system’s components. To this end, we propose an incremental experimental evaluation, aimed at (1) studying the effectiveness by which an ANN can capture the perceived interpretability for simulated users, (2) investigating how the GP’s outcome is affected across different simulated user feedback profiles, and (3) determining whether humans participants would prefer models that were generated with or without their involvement. Our results pose clarity on pros and cons of using a human-in-the-loop approach to discover interpretable ML models with GP.
- 23 February 2024
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Evolutionary Learning and Optimization
- Vol. 4 (1) , 1-30
Open AccessAn Analysis of the Ingredients for Learning Interpretable Symbolic Regression Models with Human-in-the-loop and Genetic Programming
Interpretability is a critical aspect to ensure a fair and responsible use of machine learning (ML) in high-stakes applications. Genetic programming (GP) has been used to obtain interpretable ML models because it operates at the level of functional building blocks: if these building blocks are interpretable, there is a chance that their composition (i.e., the entire ML model) is also interpretable. However, the degree to which a model is interpretable depends on the observer. Motivated by this, we study a recently-introduced human-in-the-loop system that allows the user to steer GP’s generation process to their preferences, which shall be online-learned by an artificial neural network (ANN). We focus on the generation of ML models as analytical functions (i.e., symbolic regression) as this is a key problem in interpretable ML, and propose a two-fold contribution. First, we devise more general representations for the ML models for the ANN to learn upon, to enable the application of the system to a wider range of problems. Second, we delve into a deeper analysis of the system’s components. To this end, we propose an incremental experimental evaluation, aimed at (1) studying the effectiveness by which an ANN can capture the perceived interpretability for simulated users, (2) investigating how the GP’s outcome is affected across different simulated user feedback profiles, and (3) determining whether humans participants would prefer models that were generated with or without their involvement. Our results pose clarity on pros and cons of using a human-in-the-loop approach to discover interpretable ML models with GP.