Possibilities of deep learning networks explored

The next scientific seminar of the Department No.1 of the Institute of Information Technology of ANAS on the topic "Deep Learning" was held.

The report was presented by an employee of department No. 1 Adila Imamverdiyeva. She gave information about the history of deep learning, various features of deep learning and machine learning, deep learning applications, biological neurons, direct transitions and recurrent neural networks.

Speaking about the history of deep learning, which is one of the methods of machine learning based on artificial neural networks, she said that in 1958 F. Rosenblatt proposed a single-layer perceptron, and in 1980 Kunihiko Fukushima - a neocognitron model with multi-layer artificial neural networks.

A. Imamverdiyeva talked about the differences between deep learning and machine learning, said that deep learning models automatically extract functions using large data collection and neural network architecture.

A. Imamverdiyeva provided detailed information about two classes of neural network architectures - an advanced direct neural network and a recurrent neural network. The speaker brought to the attention that information in a direct neural network is transmitted from inputs only in a straight line. She noted that in recurrent neural networks, the original data output is used as the next input.

In conclusion, A. Imamverdieva spoke about the most popular conferences and publications in this field.

© All rights reserved. Citing to www.ict.az is necessary upon using new