Vladimir Jotsov

Data science applications to improve accuracy of thermocouples

This paper considers the usage of artificial intelligence, in particular, neural networks, to correct and compensate thermocouple errors. There are the correction of the thermocouple tolerance, the error due to conversion characteristic drift under the influence of high operating temperatures as well as the compensation of the error due to acquired thermoelectric inhomogeneity of thermocouple legs proposed in this paper. The correction is carried out using individual mathematical models based on neural networks. It is proposed the neural network method for controlling a temperature field to compensate the error due to acquired thermoelectric inhomogeneity.

Innovative Issues in Intelligent Systems

This book presents a broad variety of different contemporary IT methods and applications in Intelligent Systems is displayed. Every book chapter represents a detailed, specific, far reaching and original re-search in a respective scientific and practical field. However, all of the chapters share the common point of strong similarity in a sense of being innovative, applicable and mutually compatible with each other. In other words, the methods from the different chapters can be viewed as bricks for building the next generation thinking machines as well as for other futuristic logical applications that are rapidly changing our world nowadays.

Applications of Advanced Analytics Methods in Sas Enterprise Miner

This paper considers one of the contemporary advanced analytics applications named Puzzle methods. It is studied aiming at novel results in collaborative statistical and logical research based on quantitative method applications, deep processing of accumulated knowledge, etc. It is shown that applications of intelligent technologies advance the efficiency of statistical applications. Financial and security systems (SS) have been considered as an example of difficult-to-explore areas. Original results are presented on how to build more effective logical-and-statistical applications by using novel puzzle methodologies. It is shown that all the demonstrated advantages may be successfully combined with other known methods from advanced analytics, knowledge discovery, data/web/deep data mining or other fields. Also it is shown how the considered applications enhance the quality of statistical inference, improve the human-machine interaction between the user and system and hence serve the process of sustainable improvement of the results. Applications to SAS Enterprise Miner reveal the strength of the proposed Puzzle methods.

Discrete Markov process interpretation of propositional logic

A stochastic interpretation of propositional logic formulas is introduced that uses a specific discrete Markov process with two states. The requirements for this interpretation are formulated. It is shown that the obtained from it stochastic distributions are compatible on a qualitative (probabilistic) level with the respective results of the propositional logic. Examples are presented for usage of this class of Markov processes and the ways for applying it in artificial intelligence, intelligent systems, expert systems are marked.

Human-centered methods for applications of fuzzy probabilities in intelligent systems

The paper introduces a synthetic metamethod SMM which is aimed at automation of fuzzy perceptions approval. Also the considered methods are applicable in creative processes concerning different types of intelligent systems. The metamethod is based on an analysis of incoming perception-oriented knowledge and a successive check in the domain. The proposed main methods - SMM, INCONSISTENCY, FUNNEL, and CROSSWORD - interact on a competitive principle. Two additional methods, FRONTAL and CALEIDOSCOPE, also can be put in the same group. A research is going on for the possibilities of including methods by other scientists under the general control of SMM. An evolution of the solution is used in cases of applications in environments poor of knowledge or of missing precise orientation to the final solution. The approval of fuzzy probabilities frequently needs an act of creation especially if it is applied in a knowledge-poor environment. It is shown that even in this case the creative processes are based on applications of clear and well-formalized methods included in many different sophisticated combinations.

Learning Through Constraint Applications

Solutions are considered for focusing agent attention aiming at lower computational complexity of information processing. New logical-based types of constraints and applications of Puzzle methods have been considered as a specific standard for improvement of different Data Science problems, especially via SAS Enterprise Miner tools. It is shown how the prognostic models have been improved using the considered original data-driven approaches or algorithms or by using different types of Binding, Pointing and different classical constraints. As a result, new types of non-implicative causal relations are revealed. Different modifications of proposed Puzzle methods have been researched aiming at better control of different types of constraints and elaboration of new, contemporary and more universal evolutionary applications of logical and statistical methods in one system.

Decreasing Influence of the Error Due to Acquired Inhomogeneity of Sensors by the Means of Artificial Intelligence

Sensors usually have the biggest error among all components in a measuring system. The paper considers the application of the methods of artificial intelligence, in particular, neural networks and data science applications for sensor data processing. The main attention is focused on improvement of measurement accuracy when using inaccurate sensors. The abovementioned methods illustrated on the example of improvement of measurement accuracy of the most widely used temperature sensor—the thermocouple. Neural networks and other methods of artificial intelligence ensure the improvement of accuracy of temperature measurements by an order of magnitude. However, they require considerable complication in both hardware and software.

Method for Interpretation of Functions of Propositional Logic by Specific Binary Markov Processes

The current paper proposes a method for interpretation of propositional binary logic functions using multi-binary Markov process. This allows logical concepts ‘true’ and ‘false’ to be treated as stochastic variables, and this in two ways—qualitative and quantitative. In the first case, if the probability of finding a Markov process in a definitely true state of this process is greater than 0.5, it is assumed that the Markov process is in state ‘truth.’ Otherwise the Markov process is in state ‘false.’ In quantitative terms, depending on the chosen appropriate binary matrix of transition probabilities it is possible to calculate the probability of finding the process in one of the states ‘true’ or ‘false’ for each of the steps \( n = 0 , { }1 , { }2 , { } \ldots \) of the Markov process. A single-step Markov realization is elaborated for standard logic functions of propositional logic; a series of analytical relations are formulated between the stochastic parameters of the Markov process before and after the implementation of the single-step transition. It has been proven that any logical operation can directly, uniquely, and consistently be described by a corresponding Markov process. Examples are presented and a numerical realization is realized of some functions of propositional logic by binary Markov processes.

Puzzle methods for data science applications

Applications have been considered of Puzzle methods as a specific standard for improvement of different SAS Enterprise Miner tools. It is shown how the prognostic models have been improved using the considered original data-driven approaches or algorithms or by using different types of Binding, Pointing and classical constraints. As a result, new types of non-implicative causal relations are revealed. Different modifications of the proposed Puzzle methods have been researched aiming at better control of different types of constraints and elaboration of new, contemporary and more universal evolutionary applications of logical and statistical methods in one system.

Specifics of transitions in binary Markov logic

The paper presents a research on transitive probabilities of a binary Markov logic. Analytical dependencies are obtained of such transitions. A Markov interpretation is introduced of propositional-logic formulas using only two binary Markov matrices, two rules and a suitably selected negation of binary vectors. Formulas are obtained for final transitive probabilities for each of the two selected binary Markov matrix. Numeric examples are presented for a Markov interpretation of propositional-logic logical operations.