Conditional probability density estimation using artificial neural network

This paper provides a general introduction in the field of estimation of probability density function (pdf) of data using a neural network and proposes detailed research of instruments for data preprocessing. In the first section of the research we give a review of current methods and instruments to solve the problem of pdf estimation, highlight their advantages and disadvantages and explain our decision to conduct research in this field. In the second section firstly we describe the approach which was used for estimation of pdf using neural network and give a scheme of instrument for this. Secondly we give detailed information about parts of instrument. Improvement for data preprocessing which solves problem of near-zero values and increases accuracy of instrument was proposed in the second section. In the last section we give results of experiment which approves suggested improvement and the correctness of the approach. Many aspects of pdf estimation can be improved through mathematical and analysis work. Here we present general approach and improvement for data preprocessing.