Large language models (LLMs) have recently become an intensive area of research. Their capabilities in text synthesis, summarization, and translation are well established. However, their application in classical machine learning tasks, such as classification, remains a relatively new and underexplored area.

This seminar aims to present the results of the project "Classifier for district metered areas (DMA) based on large language models (LLM)" conducted at ITAI PAS. We will present an innovative approach to model training that leverages knowledge transfer from traditional random forest classifiers. Although the idea originated from the problem of classifying water distribution systems data, which was challenged in the WaterPrime project, the seminar will demonstrate its potential in the context of classifying traditional datasets and hyperspectral images.

• Link to a preprint discussing the proposed method: http://arxiv.org/abs/2406.04926

Large language models (LLMs) have recently become an intensive area of research. Their capabilities in text synthesis, summarization, and translation are well established. However, their application in classical machine learning tasks, such as classification, remains a relatively new and underexplored area.

This seminar aims to present the results of the project "Classifier for district metered areas (DMA) based on large language models (LLM)" conducted at ITAI PAS. We will present an innovative approach to model training that leverages knowledge transfer from traditional random forest classifiers. Although the idea originated from the problem of classifying water distribution systems data, which was challenged in the WaterPrime project, the seminar will demonstrate its potential in the context of classifying traditional datasets and hyperspectral images.