Federated Learning with ResNet-18 for Medical Image Diagnosis

Abstract

Deep learning has shown promise in accurate medical image analysis, but challenges remain. Data privacy concerns hinder the availability of large, high-quality medical datasets. Traditional deep learning approaches are computationally intensive and lack efficiency. This paper proposes the use of Federated Learning (FL) with ResNet-18, a deep neural network architecture. ResNet-18 addresses gradient issues using residual blocks and skip connections. FL enables collaborative training while preserving data privacy. The training process utilizes stochastic gradient descent and techniques such as data augmentation and regularization for improved model performance.

Ähnliche Arbeiten