Development of a platform for creating neural networks in the JAVA programming language

Artificial intelligence (AI) is playing an important role in the modern world and continues to gain popularity. However, the existing services using it have a number of drawbacks and are often subject to the sanction policies, which could make them unavailable. Development of such AI systems and models requires extensive knowledge in mathematics and programming, which makes them inaccessible to the most interested users; and popular libraries for working with the neural networks could also appear out of reach for the Russian users. The paper describes the process of developing a domestic platform that makes it possible to create, configure and learn the neural networks with a possibility of their further introduction for practical purposes. The paper assumes that such a platform would be used in the future for educational purposes to familiarize new users with the basic concepts and operation principles of a fully connected neural network.

References

[1] Goodfellow Ya., Bengio I., Courville A. Deep learning. Moscow, DMK Press, 2018, 653 p. (In Russ.).

[2] Mirkes E.M. Neurocomputer. Draft standard. Novosibirsk, Nauka, Siberian Publishing Company of the Russian Academy of Sciences Publ., 1998, 190 p. (In Russ.).

[3] Khaykin S. Neural networks: a complete course. Moscow, Williams Publ., 2006, 1104 p. (In Russ.).

[4] Bloch D. Java. Effective programming. Moscow, Lori Publ., 2002, 224 p. (In Russ.).

[5] Sedgwick R., Wayne K. Algorithms in Java. Moscow, Williams Publ., 2013, 848 p. (In Russ.).