-
FM-G-CAM: A Holistic Approach for Explainable AI in Computer Vision
Authors:
Ravidu Suien Rammuni Silva,
Jordan J. Bird
Abstract:
Explainability is an aspect of modern AI that is vital for impact and usability in the real world. The main objective of this paper is to emphasise the need to understand the predictions of Computer Vision models, specifically Convolutional Neural Network (CNN) based models. Existing methods of explaining CNN predictions are mostly based on Gradient-weighted Class Activation Maps (Grad-CAM) and so…
▽ More
Explainability is an aspect of modern AI that is vital for impact and usability in the real world. The main objective of this paper is to emphasise the need to understand the predictions of Computer Vision models, specifically Convolutional Neural Network (CNN) based models. Existing methods of explaining CNN predictions are mostly based on Gradient-weighted Class Activation Maps (Grad-CAM) and solely focus on a single target class. We show that from the point of the target class selection, we make an assumption on the prediction process, hence neglecting a large portion of the predictor CNN model's thinking process. In this paper, we present an exhaustive methodology called Fused Multi-class Gradient-weighted Class Activation Map (FM-G-CAM) that considers multiple top predicted classes, which provides a holistic explanation of the predictor CNN's thinking rationale. We also provide a detailed and comprehensive mathematical and algorithmic description of our method. Furthermore, along with a concise comparison of existing methods, we compare FM-G-CAM with Grad-CAM, highlighting its benefits through real-world practical use cases. Finally, we present an open-source Python library with FM-G-CAM implementation to conveniently generate saliency maps for CNN-based model predictions.
△ Less
Submitted 13 April, 2024; v1 submitted 10 December, 2023;
originally announced December 2023.
-
AsaPy: A Python Library for Aerospace Simulation Analysis
Authors:
Joao P. A. Dantas,
Samara R. Silva,
Vitor C. F. Gomes,
Andre N. Costa,
Adrisson R. Samersla,
Diego Geraldo,
Marcos R. O. A. Maximo,
Takashi Yoneyama
Abstract:
AsaPy is a custom-made Python library designed to simplify and optimize the analysis of aerospace simulation data. Instead of introducing new methodologies, it excels in combining various established techniques, creating a unified, specialized platform. It offers a range of features, including the design of experiment methods, statistical analysis techniques, machine learning algorithms, and data…
▽ More
AsaPy is a custom-made Python library designed to simplify and optimize the analysis of aerospace simulation data. Instead of introducing new methodologies, it excels in combining various established techniques, creating a unified, specialized platform. It offers a range of features, including the design of experiment methods, statistical analysis techniques, machine learning algorithms, and data visualization tools. AsaPy's flexibility and customizability make it a viable solution for engineers and researchers who need to quickly gain insights into aerospace simulations. AsaPy is built on top of popular scientific computing libraries, ensuring high performance and scalability. In this work, we provide an overview of the key features and capabilities of AsaPy, followed by an exposition of its architecture and demonstrations of its effectiveness through some use cases applied in military operational simulations. We also evaluate how other simulation tools deal with data science, highlighting AsaPy's strengths and advantages. Finally, we discuss potential use cases and applications of AsaPy and outline future directions for the development and improvement of the library.
△ Less
Submitted 29 April, 2024; v1 submitted 11 July, 2023;
originally announced October 2023.
-
FeatGeNN: Improving Model Performance for Tabular Data with Correlation-based Feature Extraction
Authors:
Sammuel Ramos Silva,
Rodrigo Silva
Abstract:
Automated Feature Engineering (AutoFE) has become an important task for any machine learning project, as it can help improve model performance and gain more information for statistical analysis. However, most current approaches for AutoFE rely on manual feature creation or use methods that can generate a large number of features, which can be computationally intensive and lead to overfitting. To a…
▽ More
Automated Feature Engineering (AutoFE) has become an important task for any machine learning project, as it can help improve model performance and gain more information for statistical analysis. However, most current approaches for AutoFE rely on manual feature creation or use methods that can generate a large number of features, which can be computationally intensive and lead to overfitting. To address these challenges, we propose a novel convolutional method called FeatGeNN that extracts and creates new features using correlation as a pooling function. Unlike traditional pooling functions like max-pooling, correlation-based pooling considers the linear relationship between the features in the data matrix, making it more suitable for tabular data. We evaluate our method on various benchmark datasets and demonstrate that FeatGeNN outperforms existing AutoFE approaches regarding model performance. Our results suggest that correlation-based pooling can be a promising alternative to max-pooling for AutoFE in tabular data applications.
△ Less
Submitted 14 August, 2023;
originally announced August 2023.
-
Evaluation of Formal IDEs for Human-Machine Interface Design and Analysis: The Case of CIRCUS and PVSio-web
Authors:
Camille Fayollas,
Célia Martinie,
Philippe Palanque,
Paolo Masci,
Michael D. Harrison,
José C. Campos,
Saulo Rodrigues e Silva
Abstract:
Critical human-machine interfaces are present in many systems including avionics systems and medical devices. Use error is a concern in these systems both in terms of hardware panels and input devices, and the software that drives the interfaces. Guaranteeing safe usability, in terms of buttons, knobs and displays is now a key element in the overall safety of the system. New integrated developmen…
▽ More
Critical human-machine interfaces are present in many systems including avionics systems and medical devices. Use error is a concern in these systems both in terms of hardware panels and input devices, and the software that drives the interfaces. Guaranteeing safe usability, in terms of buttons, knobs and displays is now a key element in the overall safety of the system. New integrated development environments (IDEs) based on formal methods technologies have been developed by the research community to support the design and analysis of high-confidence human-machine interfaces. To date, little work has focused on the comparison of these particular types of formal IDEs. This paper compares and evaluates two state-of-the-art toolkits: CIRCUS, a model-based development and analysis tool based on Petri net extensions, and PVSio-web, a prototy** toolkit based on the PVS theorem proving system.
△ Less
Submitted 29 January, 2017;
originally announced January 2017.
