Joel Luis Carbonera
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Joel Luis Carbonera has a total of 76 co-authors. Here are the top co-authors by number of shared publications:
- Mara Abel
Federal University of Rio Grande do Sul (UFRGS)
13 shared publications Mara Abel
12 shared publications- Fabrício Henrique Rodrigues
State University of New York at Buffalo (UB)
5 shared publications - Sandro Rama Fiorini
IBM Research Brazil
5 shared publications Edson Prestes
4 shared publications- Luan Fonseca Garcia
Federal University of Rio Grande do Sul (UFRGS)
4 shared publications
Latest Publications
- 5 October 2026
- Published by Springer Nature
- p. 200-216
An Ontology Design Pattern to Represent the Selective Transitivity of Fluid Flow Connections
- 5 October 2026
- Published by Springer Nature
- p. 200-216
An Ontology Design Pattern to Represent the Selective Transitivity of Fluid Flow Connections
- 2 October 2026
- Published by Springer Nature
- p. 236-250
What Shapes Representations in Visual Models? Evidence from Architecture, Training, and Data
- 2 October 2026
- Published by Springer Nature
- p. 236-250
What Shapes Representations in Visual Models? Evidence from Architecture, Training, and Data
- 30 September 2026
- Published by Springer Nature
- p. 626-640
Adaptive Multi-objective Reinforcement Learning for Context-Conditioned Water Temperature Control
- 30 September 2026
- Published by Springer Nature
- p. 626-640
Adaptive Multi-objective Reinforcement Learning for Context-Conditioned Water Temperature Control
- Open Access
- 26 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 49121-49140
Digital Twins Monitoring: A Methodology for Generating and Validating Synthetic Failure Data for PHM
A digital twin is a software system connected to a physical plant for continuous monitoring and control of its behavior. The permanent data sensing provides insights for the maintenance guide, also known as Prognostics and Health Management (PHM). However, in most industrial installations, failure episodes are rare due to the operational nature of machines, and plant sensors capture limited data to support analysis and decision-making. Therefore, the limited availability of failure data is a significant challenge in modeling, identifying, and predicting failures in real scenarios. Advances in computing power, simulation, sensing, and networking technologies for digital twins now offer a novel approach: using data augmentation by introducing failures into the digital replica of the real asset to train data-driven PHM models. This paper proposes a data processing pipeline for synthesizing and validating failure data for PHM purposes using a small amount of sensed data, a simulator, and domain knowledge. Our methodology creates a synthetic failure dataset, validated against real data, that enables us to explore machine learning for failure prediction and cause diagnosis. In our experiment, we modeled a smart petroleum well using a commercial computational fluid dynamics simulator, inducing failures by changing the equipment’s expected behavior to produce synthetic failure data.We assessed the quality of this synthetic data by training machine learning algorithms on it, testing it with a labeled failure process dataset from a petroleum plant, and applying fidelity metrics to identify necessary improvements. Finally, we compared our methodology’s performance with that of generative adversarial networks (GANs) for generating synthetic failure data. Our results show that our methodology outperforms GANs at generating failure data without requiring a large labeled dataset.
- 26 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 49121-49140
Open AccessDigital Twins Monitoring: A Methodology for Generating and Validating Synthetic Failure Data for PHM
A digital twin is a software system connected to a physical plant for continuous monitoring and control of its behavior. The permanent data sensing provides insights for the maintenance guide, also known as Prognostics and Health Management (PHM). However, in most industrial installations, failure episodes are rare due to the operational nature of machines, and plant sensors capture limited data to support analysis and decision-making. Therefore, the limited availability of failure data is a significant challenge in modeling, identifying, and predicting failures in real scenarios. Advances in computing power, simulation, sensing, and networking technologies for digital twins now offer a novel approach: using data augmentation by introducing failures into the digital replica of the real asset to train data-driven PHM models. This paper proposes a data processing pipeline for synthesizing and validating failure data for PHM purposes using a small amount of sensed data, a simulator, and domain knowledge. Our methodology creates a synthetic failure dataset, validated against real data, that enables us to explore machine learning for failure prediction and cause diagnosis. In our experiment, we modeled a smart petroleum well using a commercial computational fluid dynamics simulator, inducing failures by changing the equipment’s expected behavior to produce synthetic failure data.We assessed the quality of this synthetic data by training machine learning algorithms on it, testing it with a labeled failure process dataset from a petroleum plant, and applying fidelity metrics to identify necessary improvements. Finally, we compared our methodology’s performance with that of generative adversarial networks (GANs) for generating synthetic failure data. Our results show that our methodology outperforms GANs at generating failure data without requiring a large labeled dataset.
- 12 November 2025
- Published by Sociedade Brasileira de Computação
- p. 152-155
Seleção de amostras baseada em transfer learning e autoaprendizagem para tarefas zero-shot
Em ambientes com escassez de dados visuais rotulados, como no contexto médico ou geológico, modelos de aprendizado profundo sofrem com problemas de desempenho. Nesse contexto, desenvolvemos um pipeline utilizando transfer learning e auto-aprendizagem para viabilizar o treinamento de um classificador, podendo ser utilizado para diferentes tarefas zero-shot como classificação de imagens ou detecção de amostras de fora da distribuição (OOD). Avaliações experimentais demonstram que nossa abordagem supera o estado-da-arte dessas tarefas.
- 12 November 2025
- Published by Sociedade Brasileira de Computação
- p. 152-155
Seleção de amostras baseada em transfer learning e autoaprendizagem para tarefas zero-shot
Em ambientes com escassez de dados visuais rotulados, como no contexto médico ou geológico, modelos de aprendizado profundo sofrem com problemas de desempenho. Nesse contexto, desenvolvemos um pipeline utilizando transfer learning e auto-aprendizagem para viabilizar o treinamento de um classificador, podendo ser utilizado para diferentes tarefas zero-shot como classificação de imagens ou detecção de amostras de fora da distribuição (OOD). Avaliações experimentais demonstram que nossa abordagem supera o estado-da-arte dessas tarefas.
- Open Access
No Labels Needed: Zero-Shot Image Classification with Collaborative Self-Learning
While deep learning, including Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), has significantly advanced classification performance, its typical reliance on extensive annotated datasets presents a major obstacle in many practical scenarios where such data is scarce. Vision-language models (VLMs) and transfer learning with pre-trained visual models appear as promising techniques to deal with this problem. This paper proposes a novel zero-shot image classification framework that combines a VLM and a pre-trained visual model within a self-learning cycle. Requiring only the set of class names and no labeled training data, our method utilizes a confidence-based pseudo-labeling strategy to train a lightweight classifier directly on the test data, enabling dynamic adaptation. The VLM identifies high-confidence samples, and the pre-trained visual model enhances their visual representations. These enhanced features then iteratively train the classifier, allowing the system to capture complementary semantic and visual cues without supervision. Notably, our approach avoids VLM fine-tuning and the use of large language models, relying on the visual-only model to reduce the dependence on semantic representation. Experimental evaluations on ten diverse datasets demonstrate that our approach outperforms the baseline zero-shot method.
Open AccessNo Labels Needed: Zero-Shot Image Classification with Collaborative Self-Learning
While deep learning, including Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), has significantly advanced classification performance, its typical reliance on extensive annotated datasets presents a major obstacle in many practical scenarios where such data is scarce. Vision-language models (VLMs) and transfer learning with pre-trained visual models appear as promising techniques to deal with this problem. This paper proposes a novel zero-shot image classification framework that combines a VLM and a pre-trained visual model within a self-learning cycle. Requiring only the set of class names and no labeled training data, our method utilizes a confidence-based pseudo-labeling strategy to train a lightweight classifier directly on the test data, enabling dynamic adaptation. The VLM identifies high-confidence samples, and the pre-trained visual model enhances their visual representations. These enhanced features then iteratively train the classifier, allowing the system to capture complementary semantic and visual cues without supervision. Notably, our approach avoids VLM fine-tuning and the use of large language models, relying on the visual-only model to reduce the dependence on semantic representation. Experimental evaluations on ten diverse datasets demonstrate that our approach outperforms the baseline zero-shot method.
- 22 July 2025
- Published by Elsevier in Knowledge-Based Systems
- Vol. 327
Using natural language definitions and language models for relationship classification
- 22 July 2025
- Published by Elsevier in Knowledge-Based Systems
- Vol. 327
Using natural language definitions and language models for relationship classification
- Open Access
An approach based on class activation maps for investigating the effects of data augmentation on neural networks for image classification
Neural networks have become increasingly popular in the last few years as an effective tool for the task of image classification due to the impressive performance they have achieved on this task. In image classification tasks, it is common to use data augmentation strategies to increase the robustness of trained networks to changes in the input images and to avoid overfitting. Although data augmentation is a widely adopted technique, the literature lacks a body of research analyzing the effects data augmentation methods have on the patterns learned by neural network models working on complex datasets. The primary objective of this work is to propose a methodology and set of metrics that may allow a quantitative approach to analyzing the effects of data augmentation in convolutional networks applied to image classification. An important tool used in the proposed approach lies in the concept of class activation maps for said models, which allow us to identify and measure the importance these models assign to each individual pixel in an image when executing the classification task. From these maps, we may then extract metrics over the similarities and differences between maps generated by these models trained on a given dataset with different data augmentation strategies. Experiments made using this methodology suggest that the effects of these data augmentation techniques not only can be analyzed in this way but also allow us to identify different impact profiles over the trained models.
Open AccessAn approach based on class activation maps for investigating the effects of data augmentation on neural networks for image classification
Neural networks have become increasingly popular in the last few years as an effective tool for the task of image classification due to the impressive performance they have achieved on this task. In image classification tasks, it is common to use data augmentation strategies to increase the robustness of trained networks to changes in the input images and to avoid overfitting. Although data augmentation is a widely adopted technique, the literature lacks a body of research analyzing the effects data augmentation methods have on the patterns learned by neural network models working on complex datasets. The primary objective of this work is to propose a methodology and set of metrics that may allow a quantitative approach to analyzing the effects of data augmentation in convolutional networks applied to image classification. An important tool used in the proposed approach lies in the concept of class activation maps for said models, which allow us to identify and measure the importance these models assign to each individual pixel in an image when executing the classification task. From these maps, we may then extract metrics over the similarities and differences between maps generated by these models trained on a given dataset with different data augmentation strategies. Experiments made using this methodology suggest that the effects of these data augmentation techniques not only can be analyzed in this way but also allow us to identify different impact profiles over the trained models.
- 29 November 2024
- Published by Elsevier in Future Generation Computer Systems
- Vol. 166
A Framework for testing Federated Learning algorithms using an edge-like environment
- 29 November 2024
- Published by Elsevier in Future Generation Computer Systems
- Vol. 166
A Framework for testing Federated Learning algorithms using an edge-like environment
- Open Access
Accelerating prototype selection with spatial abstraction
The increasing digitalization in industry and society leads to a growing abundance of data available to be processed and exploited. However, the high volume of data requires considerable computational resources for applying machine learning approaches. Prototype selection techniques have been applied to reduce the requirements of computational resources that are needed by these techniques. In this paper, we propose an approach for speeding up existing prototype selection techniques. It builds an abstract representation of the dataset, using the notion of spatial partition. The second step uses this abstract representation to prune the search space efficiently and select a set of candidate prototypes. After, some conventional prototype selection algorithms can be applied to the candidates selected by our approach. Our approach was integrated with five conventional prototype selection algorithms and tested on 14 widely recognized datasets used in classification tasks. The performance of the modified algorithms was compared to that of their original versions in terms of accuracy and reduction rate. The experimental results demonstrate that, overall, our proposed approach maintains accuracy while enhancing the reduction rate of the original prototype selection algorithms and simultaneously reducing their execution times.
Open AccessAccelerating prototype selection with spatial abstraction
The increasing digitalization in industry and society leads to a growing abundance of data available to be processed and exploited. However, the high volume of data requires considerable computational resources for applying machine learning approaches. Prototype selection techniques have been applied to reduce the requirements of computational resources that are needed by these techniques. In this paper, we propose an approach for speeding up existing prototype selection techniques. It builds an abstract representation of the dataset, using the notion of spatial partition. The second step uses this abstract representation to prune the search space efficiently and select a set of candidate prototypes. After, some conventional prototype selection algorithms can be applied to the candidates selected by our approach. Our approach was integrated with five conventional prototype selection algorithms and tested on 14 widely recognized datasets used in classification tasks. The performance of the modified algorithms was compared to that of their original versions in terms of accuracy and reduction rate. The experimental results demonstrate that, overall, our proposed approach maintains accuracy while enhancing the reduction rate of the original prototype selection algorithms and simultaneously reducing their execution times.