Mathematical machine usage in medical information systems

Biosystem of the human body is viewed as a whole. First of all adequate mathematical machine selection and class of biosystems needs to be assigned for creation of mathematical model of biological system. Biosystem has two types of appoach. One of them is supposed to be a simple approach, the other is likely to be very complex – indexed approach. Different biosystems with determination properties are usually described by differential and integral equations, linear and nonlinear algebra. In some cases, algebraic polynoms with timed argument are used for presenting determined biosystem dynamics. Adequate mathematical modeling machine, probability theory, Markov and random processes theory and the laws are applied for the description of likely characterized biosystems. Key words: biosystem, biocybernetic issues, differential and integral equations, mathematical model, Markov chains, Bayes method, artifical neural networks

UNDERSTANDING AND STUDY OF WEIGHT INITIALIZATION IN ARTIFICAL NEURAL NETWORKS WITH BACK PROPAGATION ALGORITHM

There are various important choices that need to be assumed when building and training a neural network. One has to determine which loss function to be used, how many layers to be include, what stride and kernel size to use for each layer, which optimization algorithm is best suited for the network and so on. Assuming all the above condition, it decided to initialize the neural network training by different weight initialization techniques. This process is carried out in affiliation or with respect to with random learning rate so that we can get better result. We have calculated the mean test error for newly proposed paradigm and traditional approach. The newly proposed paradigm Xavier Weight Initialization less error in comparison to the traditional approach of Uniform and Gaussian Weight initialization (Random Initialization).

Sintering temperature influence on grains function distribution by neural network application

Artificial neural networks application in science and techonology has begun during twentieth century. This biophysical and biomimetic phenomena is based on extensive research which have led to understanding how neural as a living organism nerve system basic element processes signals by a simple algorithm; the input signals are massively parallel processed, and the output presents the superposition of all parallel processed signals. Artifical neural networks which are based on these principles are useful for solving various problems as pattern recognition, clustering, functional optimization. This research analyzed termophysical parameters at samples based on Murata powders and consolidated by sintering process. Among different physical properties we applied out neural network approach on grain sizes distribution as a function of sintering temperature (T), (from 1190-1370?C). In this paper, we continue to apply neural networks to prognose structural and thermophysical parameters. For consolidation sintering process is very important to prognoze and design many parameters but especially thermal like temperature, to avoid long and even wrong experiments which are waisting the time and materials and energy as well. By this ANN method we indeed provide the most efficient procedure in projecting the mentioned parameters and provide successful ceremics samples production. This is very helpful in prediction and designing the microstructure parameters important for advance microelectronic further miniturisation development. This is a quite original novelty for real microstructure projecting especially on the phenomena within the thin films coating around the grains what opens new prospectives in advance fractal microelectronics.