Improving ANN Efficiency using PostgreSQL

The paper describes an approach that implements the interaction of an artificial neural network (ANN) with a database to speed up the main stages of the ANN operation. The developed approach is based on the PostgreSQL database management system (DBMS) using the PL/Python extension and the TensorFlow framework and allows building, training, saving and testing neural networks, adjusting the necessary parameters at each stage. The results of the work were tested on the example of an ANN for recognizing wheat diseases and demonstrated their efficiency in terms of reducing the operating time of the ANN compared to the standard Python approach without using a database.