Automated Deep Learning

Automated Deep Learning (AutoDL) aims to automate the entire deep learning workflow, minimizing human intervention in tasks like model architecture design, hyperparameter optimization, and feature selection. Current research focuses on developing AutoDL frameworks for diverse applications, including time series forecasting (e.g., energy load and wind power prediction), sequential recommendation systems, and medical image analysis (e.g., cancer diagnosis from microscopy images). These advancements promise to improve the efficiency, accuracy, and accessibility of deep learning across various scientific disciplines and practical domains, particularly where expert knowledge or large labeled datasets are scarce.