Paapa Kwesi Quansah

Preserving Crop Health: Harnessing the Power of Federated Continual Learning for Plant Disease Detection.

Description: In this paper, I explore an innovative approach in agriculture, focusing on the integration of Federated Continual Learning (FCL). This method combines the strengths of decentralized model training and continuous adaptive learning. My aim is to significantly enhance the accuracy and generalization of plant disease detection across diverse crops and environments. This research was born out of a need to minimize crop losses and strengthen food security by revolutionizing disease detection methods.

Short-Term Load Forecasting Using a PSO-Optimized Multi-Head Attention-Augmented CNN-LSTM Network (PSO-A2C-LNet).

Description: In the presented paper, I address the critical challenge of short-term load forecasting in power systems, a task marked by its non-linear and dynamic nature. Recognizing the limitations of current deep learning methods, such as hyperparameter sensitivity, lack of interpretability, and high computational demands, I introduce an innovative solution. This solution leverages Particle-Swarm Optimization to autonomously optimize hyperparameters, incorporates a Multi-Head Attention mechanism for enhanced feature discernment, and employs a streamlined framework to boost computational efficiency. The effectiveness of this method is validated through extensive testing on a real electricity demand dataset, demonstrating its superior accuracy, robustness, and efficiency.

FENNs: A Resource-Efficient, Adaptive, Privacy-Preserving Decentralized Learning Framework

Description: In this research, we introduce Federated Ephemeral Neural Networks (FENNs), an innovative approach addressing the dual challenges of resource intensity and data privacy in deep learning. FENNs integrate ephemeral neural networks (ENNs) for dynamic architectural adaptability with federated learning's privacy features, optimizing network structures for various tasks within decentralized, resource-constrained environments. Our testing on limited-resource devices in federated settings, alongside a novel metric for resource-constrained learning, demonstrates FENNs' effectiveness and potential in advancing edge computing and decentralized AI applications.

Under Review, 2023

An Improved Multi-Objective Grey Wolf Optimization Algorithm for Parameter Tuning

Description: In this research, I explore advancements in the Grey Wolf Optimization (GWO) algorithm, with a significant focus on optimizing power systems in renewable energy settings. This study enhances the traditional GWO approach to better handle the complexities of multi-objective optimization tasks, particularly in the dynamic field of renewable energy. The improvements in the algorithm are geared towards achieving more efficient parameter tuning, which is crucial for optimizing the performance of renewable energy systems. By improving convergence speed and accuracy, the algorithm adeptly addresses the intricate balance of various objectives in power systems, making it an efficient tool for renewable energy optimization and beyond.

Manuscript Under Preparation, 2023