Stage 5

Development and research of methods and algorithms for multi-modal training of composite models, including training on text, images, time series, and tabular data; and development of effective tools for their application within the framework of the framework.
Designing and investigating methods and algorithms for multi-modal learning of composite models, including multi-modal composite models; and developing effective tools for their application within the framework.
Prototyping acceleration of computation and data handling by switching to technologies that enable optimized data processing (e.g., Rapids, Apache Arrow), and support for hybrid (CPU + GPU) modes
Development of optimized templates for popular types of tasks and data (optimization for work with tabular data, text, images, time series, etc.).
Support of extended functionality of "raw" data preprocessing modules in terms of integration with modern libraries for feature selection, data conversion, anomaly search, etc. for multi-modal data.
Development of a set of ready-made examples of FEDOT usage for different subject areas: oil, geo-data, finance, etc.