-
Novel Saliency Analysis for the Forward Forward Algorithm
Authors:
Mitra Bakhshi
Abstract:
Incorporating the Forward Forward algorithm into neural network training represents a transformative shift from traditional methods, introducing a dual forward mechanism that streamlines the learning process by bypassing the complexities of derivative propagation. This method is noted for its simplicity and efficiency and involves executing two forward passes the first with actual data to promote…
▽ More
Incorporating the Forward Forward algorithm into neural network training represents a transformative shift from traditional methods, introducing a dual forward mechanism that streamlines the learning process by bypassing the complexities of derivative propagation. This method is noted for its simplicity and efficiency and involves executing two forward passes the first with actual data to promote positive reinforcement, and the second with synthetically generated negative data to enable discriminative learning. Our experiments confirm that the Forward Forward algorithm is not merely an experimental novelty but a viable training strategy that competes robustly with conventional multi layer perceptron (MLP) architectures. To overcome the limitations inherent in traditional saliency techniques, which predominantly rely on gradient based methods, we developed a bespoke saliency algorithm specifically tailored for the Forward Forward framework. This innovative algorithm enhances the intuitive understanding of feature importance and network decision-making, providing clear visualizations of the data features most influential in model predictions. By leveraging this specialized saliency method, we gain deeper insights into the internal workings of the model, significantly enhancing our interpretative capabilities beyond those offered by standard approaches. Our evaluations, utilizing the MNIST and Fashion MNIST datasets, demonstrate that our method performs comparably to traditional MLP-based models.
△ Less
Submitted 18 September, 2024;
originally announced September 2024.
-
Optimizing Underwater IoT Routing with Multi-Criteria Decision Making and Uncertainty Weights
Authors:
Ali Karkehabadi,
Mitra Bakhshi,
Seyed Behnam Razavian
Abstract:
Effective data routing is vital in the Internet of Things (IoT) paradigm, especially in underwater mobile sensor networks where inefficiency can lead to significant resource consumption. This article presents an innovative method designed to enhance network performance and reduce resource usage, while also accurately determining component weights in these networks, ensuring quality service. Buildi…
▽ More
Effective data routing is vital in the Internet of Things (IoT) paradigm, especially in underwater mobile sensor networks where inefficiency can lead to significant resource consumption. This article presents an innovative method designed to enhance network performance and reduce resource usage, while also accurately determining component weights in these networks, ensuring quality service. Building upon previous research on multi-criteria decision-making systems in coastal RPL networks, our method involves key adaptations for underwater environments. It integrates comprehensive network features to identify the optimal parent node for each sensor, employing the fuzzy SWARA decision-making approach under uncertain conditions. This method takes into account various factors including hops, energy, ARSSI rate, delay, ETX, link delivery rate, and depth to determine the most effective parent node assignment. Through simulation, our approach demonstrates marked improvements in network performance compared to existing solutions. These advancements are significant, offering a new direction in enhancing underwater IoT communications and suggesting wider applications for IoT systems facing similar challenges.
△ Less
Submitted 19 May, 2024;
originally announced May 2024.
-
User-Centric Optimization for Constraint Web Service Composition using a Fuzzy-guided Genetic Algorithm System
Authors:
Mahdi Bakhshi,
Mohsen Hashemi
Abstract:
Service-Oriented Applications (SOA) are being regarded as the main pragmatic solution for distributed environments. In such systems, however each service responds the user request independently, it is essential to compose them for delivering a compound value-added service. Since, there may be a number of compositions to create the requested service, it is important to find one which its properties…
▽ More
Service-Oriented Applications (SOA) are being regarded as the main pragmatic solution for distributed environments. In such systems, however each service responds the user request independently, it is essential to compose them for delivering a compound value-added service. Since, there may be a number of compositions to create the requested service, it is important to find one which its properties are close to user's desires and meet some non-functional constraints and optimize criteria such as overall cost or response time. In this paper, a user-centric approach is presented for evaluating the service compositions which attempts to obtain the user desires. This approach uses fuzzy logic in order to inference based on quality criteria ranked by user and Genetic Algorithms to optimize the QoS-aware composition problem. Results show that the Fuzzy-based Genetic algorithm system enables user to participate in the process of web service composition easier and more efficient.
△ Less
Submitted 12 October, 2012;
originally announced October 2012.
