AIONet: A HYBRID ALGORITHM FOR AUTOMATED MODEL AND DATASET SELECTION BASED ON DECISION TREES AND LARGE LANGUAGE MODELS

Abstract

The article examines an adaptive algorithm for automated selection and training of machine learning models, AIONet (Adaptive Intelligent Optimization Network), based on the integration of decision tree methods and large language models (LLMs). A comparative analysis is conducted of the capabilities of classical decision tree algorithms and modern LLMs in the task of identifying suitable models and datasets for automated neural network training. A hypothesis is formulated that combining a decision tree as a structure for primary logical selection with an LLM as a context-dependent intelligent module can improve the accuracy and efficiency of model and dataset selection compared to using each approach independently. The key features of the AIONet algorithm’s operation are presented, as well as its potential for application in automated machine learning systems.