Main Concepts

The main framework concepts are as follows:

• Flexibility. FEDOT can be used to automate the construction of solutions for various problems, data types (texts, images, tables), and models;

• Extensibility. Pipeline optimization algorithms are data- and task-independent, yet you can use special strategies for specific tasks or data types (time-series forecasting, NLP, tabular data, etc.) to increase the efficiency;

• Integrability. FEDOT supports widely used ML libraries (Scikit-learn, CatBoost, XGBoost, etc.) and allows you to integrate custom ones;

• Tuningability. Various hyper-parameters tuning methods are supported including models’ custom evaluation metrics and search spaces;

• Versatility. FEDOT is not limited to specific modeling tasks, for example, it can be used in ODE or PDE;

• Reproducibility. Resulting pipelines can be exported separately as JSON or together with your input data as ZIP archive for experiments reproducibility;

• Customizability. FEDOT allows managing models complexity and thereby achieving desired quality.

The comparison of fedot with main existing AutoML tools is provided below: