Simulating learning methodology (SLeM): an approach to machine learning automation.

The article discusses the concept of Self-Learning Meta (SLeM) and its applications in Auto Machine Learning (AutoML). SLeM is a framework that aims to automate various components of the machine learning process, such as data selection, model adjustment, loss setting, and algorithm designing. The framework utilizes meta-data or meta-knowledge to evaluate and predict the optimal learning methodology for different tasks. The text presents several SLeM algorithms, including data auto-selection, model auto-adjustment, loss auto-setting, and algorithm auto-designing. These algorithms have been implemented and released on an open-source platform. The text also highlights the challenges and future research directions in developing more automated and comprehensive SLeM algorithms. The work was supported by funding from the National Key Research and Development Program of China and the National Natural Science Foundation of China. [Extracted from the article]