Kalin Stoyanov

University of Chemical Technology and Metallurgy | UCTM · Department of Automation

The process of training of Artificial Neural Networks essentially is optimization of the values of the weights $ w_{pq} $ associated with the arcs, connecting the nodes of the layers. This is a process of minimization of the Loss function (maximization of Accuracy function). During the training, the training data set recursively is utilized at subs...

A training discipline "Basics if the Artificial Intelligence ", taught in UCTM, Sofia is presented. The discipline is developed according to the requirements of ACM as well the methodises presented by Russel and Norvig in their seminal book “Artificial Intelligence: A modern approach, 3 ed.”. As a tool for illustration of the AI algorithms to be st...

The scanning of factor spaces (both process and mixture), usually is referred to as a technique for creation of a set of points. This set can be applied as the domain, where the values of some response function are calculated. Alternatively this can be the set of candidate points, used for searching of optimal experimental designs. In many practica...

Clinically, there is an urgent need to identify new therapeutic strategies for selectively treating cancer cells. One of the directions in this research is the development of biocompatible therapeutics that selectively target cancer cells. Here, we show that novel aminated graphene oxide (haGO-NH2) nanoparticles demonstrate increased toxicity towar...

The goal is to represent a multivariate polynomial in a vector space. This will allow us to apply the linear algebra tools in searching the proper structure of the polynomial. Typical application is in the stepwise method for model selection in regression analysis ([4]), but it can be used in other areas, where finding of the best structure of a po...

This chapter of the book "Practical guide to chemometrics", 2nd ed. Edited by Paul Gemperline, is intended to describe the basic methods applied in the construction of experimental designs. The advanrages of the optimal designs against the one-variable-at-a-time are underlined. Different types of optimal designs and their areas of usfulnes are des...

The book contains tables of ready constructed optimal designs. Each design is compromise optimal for two regression models. The researcher is free to choose the number of the experiments. The book contains designs for up to 10 mixture and process variables for different types of polynomial models.