A model of hierarchical temporal memory for identification of text documents ranking system is proposed. An approach to the evaluating of the model's parameters is proposed and an evaluation of training time is given. Tests were performed on data, received by modeling OkapiBm25 algorithm, applied to the collection of text documents of ROMIP seminar. The obtained results allow us to conclude that the model can solve the identification problems. Keywords-identification system, text document ranking algorithm, hierarchical temporal memory.

No abstract

Neural network models for the analysis of the document ranking algorithms are proposed. The models use the Kohonen neural network and a multilayer perceptron. These models were verified using test data, and their application features were revealed depending on the input data.

The problems of programming memristor arrays (memristor crossbars) are considered. An estimate for the pulse width to set the desired memristor resistance (memristance) value is obtained. The implementation of the Winner-Take-All (WTA) neural network on the memristor crossbar and the NMOS transistors for binary images recognition is proposed. The proposed WTA network implementation by simulation on the LTspice IV software was approved.