YSU

YEREVAN STATE UNIVERSITY

FACULTY OF ECONOMICS AND MANAGEMENT

SCIENTIFIC-ANALYTICAL CENTRE OF CONSTITUTIONAL ECONOMICS

THE APPLICATION OF NEURAL NETWORK IN OBJECTS'S CLUSTERS PART 2

The scientific seminar conducted by Gayane Ghukasyan, Associate Professor of Mathematical Modeling was continued at the YSU Scientific-Analytical Center of Constitutional Economics.

Continuing the topic, the speaker presented the difference between the artificial neural networks and the architecture machines of Von Neumann.

Computer systems based on artificial neural networks are endowed with a number of qualities lacking in the architecture machines of Von Neumann (but typical of the human brain).

Among these can be mentioned the massive parallels, distributed performance of information and calculations. The speaker also presented the "Dead neurons issue".

During the competition, some neurons  having vectors of initial weights significantly removed from input vectors never win the competition, no matter how long the training is.

The learning algorithm is modified to allow these neurons to participate in the competition. The winner neuron is determined by the WTA algorithm.