Kit4DL: Towards fast prototyping and experimentation in machine learning and deep learning

Abstract

Artificial neural networks, deep learning and machine learning are versatile data-driven tools widely applied in different disciplines such as finance, image and voice recognition, and earth science. For scientists and enthusiasts (including those not very experienced with programming), there is a need for easy-to-use and fast-to-setup tools that enable users to prototype and focus on the research part quickly rather than spending time on data preparation, on extracting features and setup multiple experiments for training and validating models. In this paper, we introduce Kit4DL, which is a Python package to speed up the experimentation process of machine- and deep-learning by using just a single TOML configuration file, allowing a user to set up all aspects involved in training and validation. Though simple to use in its default mode, the proposed package enables high customisation possibilities for more experienced users. Kit4DL streamlines the deep learning development process by simplifying the creation of the entire training, validation, and testing loop. Users only need to implement a few core methods outlined in a provided configuration file, significantly reducing development time compared to traditional approaches requiring from a user to implement all procedures him/herself. Additionally, Kit4DL facilitates code reusability by allowing researchers to leverage the same codebase across multiple experiments, reducing redundancy and streamlining the experimentation process.

Keywords

• Deep learning
• Machine learning
• Prototyping
• Python framework