Universal Intellectual System

1. Mark D. Kats, Dr. Sc., Academician of Ukrainian Process Engineering Academy and Ukrainian Academy of Ecological Sciences, Corresponding Member of International Academy of Computer Sciences and Systems, specialist in applied mathematics. Date of birth: January 12, 1937. Address: Donetskaya St. 37-24, Severodonetsk, Lugansk Reg., 349940, Ukraine, E-mail: (Kats M.D.),Tel.: (380-6452)- 3-08-75.

2. Name of the project.

«Development of the UNIVERSAL INTELLECTUAL SYSTEM for generating NEW KNOWLEDGE in different SUBJECT FIELDS as a base for creation of LOCAL INTELLECTUAL SYSTEMS which use this knowledge to predict the behavior, diagnostics and optimization of SPECIFIC COMPLEX TECHNICAL, BIOLOGICAL and SOCIAL OBJECTS (SYSTEMS)».

3. The offer is related to developments in the field of artificial intellectual. It is a result of strategic research aimed at creation of basis, i.e. development of methodological foundations, mathematical methods of solution of ill-conditioned (incorrectly formulated) tasks and software for their realization; development of an universal structure allowing to transfer input model into LOCAL INTELLECTUAL SYSTEM able to solve such tasks after minor text corrections.

4. Description of the offer.

METHODOLOGY OF STUDYING COMPLEX OBJECTS related to class of "large-scale" systems has been developed, tested and experimentally proved in solving a number of scientific, research and applied tasks in the field of chemistry, biology, medicine, sociology, and process technologies for chemical pharmaceutics, biotechnics, metallurgy .

Mathematical basis of the methodology consists in new methods of mathematical simulation, namely, the methods of mosaic portrait and suboptimization, i.e. logical and situational programming.

Method of mosaic portrait using formalized procedures without presence of an expert in the field under investigation and basing wholly on experimental information obtained by observing the object, system or phenomena of different nature, allows to create a mathematical model containing a great volume of new, non-trivial and formerly unknown information.

Initial information for mosaic model creation is the table of experimental data each line of which contains input and output variables in one realization of the object under investigation.

Method of mosaic portrait consists in transferring input and output variables from numeric scale to denomination scale; organization of polynomial formalized procedure in terms of computer time allowing to find combinations of denominations of input variable subranges which in the table of experimental data can be found only in one class of target function and are absent in all the experiments related to other classes. These combinations are interpreted as formal hypothesis consistent with the experimental data explaining the reason of difference of the classes.

The obtained hypotheses are easily transferred in terms of the subject field under investigation. Informational interpretation of unknown hypotheses by specialists allows to substantially increase knowledge on behavior of the object under investigation.

Mosaic model of the process under investigation serves as the initial material for its suboptimization using logical and situational programming.

Due to discrete structure of the mosaic model the task of suboptimization is assigned to class of discrete programming. It is known that no general methods are available for this purpose besides exhaustive search connected with exponential dependence of computer time for optimization on dimension of the vector of input variables which means that the tasks with larger dimensions practically cannot be solved. Unlike the known methods of discrete programming the computer time to realize methods of logical and situational programming polynomially depends on dimension of vector of input variables, thus it cannot be realized for reasonable time even if the number of variables is > 1000.

Methods of logical and situational programming are based on well known assumption of logical algebra about constituent - comprising statement, i.e. such statement is true only when it is obtained by pooling simple true statements and it is false when it comprises even as little as one false statement.

When hypotheses of "good"-class mosaic model are interpreted as simple true statements and those of "bad"- class as simple false ones , then the constituent-comprising statement of n range (n-dimension of vector of input parameters) obtained by pooling "good"-class hypotheses and comprising not a single combination of subrange codes corresponding to "bad"-class, can be considered as description of desired space domain of input variables which allows to find the specified values of generalized criterion wherein input variables of system under investigation are convoluted.

This methodology allows to create principally novel lead in informative technologies, namely, the industry generating new knowledge in any subject field.

5. Advantages comparing to analogs.

Methodology used in creation of intellectual systems is the result of strategic researches aimed at developing formal methods for investigation of "large-scale" systems using experimental data.

For a number of interesting research and applied tasks connected with investigation of objects belonging to "large-scale" systems, i.e. having high dimensions of input and output vectors, complex non-linear input/output relationships, etc. there are no mathematical means available suitable for their identification (creation of the model basing on experimental data).

It is connected with the fact that three main tasks of identification of "large-scale" systems, i.e. formal structural identification (definition of general form of the model, list of significant variables and their interaction); correct parametric identification (estimation of coefficient values of each structural element) ; formal convolution of vector of output variables into generalized criterion have not been solved yet.

Principal difficulties arising in course of creating mathematical models of real objects and their low effectiveness in solving acute scientific and practical problems resulted in searching new alternative approaches.

One of them widespread nowadays is development of expert systems. Expert system is a computation system comprising formalized knowledge of specialists in any specified field and able to make expert decisions like a man.

Effectiveness of expert system depends primarily on quantity and quality of information included in its knowledge base which is formed by subjective limited knowledge of specialists who cannot formalize it into well-defined rules , moreover many of them do not realize which rules they are following.

Besides, phsyco-physical peculiarities of perception are very individual. Any specialist understands easily 1 input/ 1 output variable relationship. A qualified specialist understands the pattern of relationship of 2 input/1 output variables. And only genius can describe the relationship of 3 input/ 1 output variables. That is why all known nature laws describe the relationship of maximum 3 independent variables.

An expert faces the same and even greater difficulties when describing the influence of more than 2 input parameters on complex criterion including more than 2 output variables .

Practical application of UNIVERSAL INTELLECTUAL SYSTEM has shown that hypotheses with 3 to 5 and more variables are rather the rule than an exception in creation of models of differential diagnostics "Gastric Ulcer - Gastric Cancer"; prediction of myocardial infarction complications in acute stage; relationship between structure and properties of dispersed monoazodye stuffs; dependence of output variables (output, efficiency, product quality characteristics, etc.) from process operating conditions in biotechnological, chemico-pharmaceutical, etc. production plants. It was hyposeses with 3 or more variables that after information interpretation turned out to be unknown to experts.

It may occur that after having spent great expenditures on various expert systems the analysis of their effectiveness would show that was known a priori:

- When solving relatively simple tasks which can be solved by an expert with the help of non-formalized methods, the effectiveness of decision made by expert system and by an expert is close and sufficient.

- When solving complex and important tasks, e.g. differential diagnostics of difficult for distinguishing diseases, etc. The effectiveness of decision (precise diagnosis) made by expert system and by an expert will be close and insufficient.

As practically all known at present EXPERT SYSTEMS and INTELLECTUAL SYSTEMS generating new knowledge with the help of formalized procedures employ the same principle of logical algebra IF - THEN for hypotheses, with expert systems accumulating all the knowledge available in the given field, then having the required volume of initial experimental information we obtain the following:

- intersection of statement sets of expert and intelligent systems will derive the subset of trivial statements;

- logical difference between statement sets of expert and intelligent systems is the subset of false hypotheses;

- logical difference between statement sets of intelligent and expert systems is the subset of hypothesis containing new non-trivial information of generalities of the object (system) under investigation.

6. Potential users.

The UNIVERSAL INTELLECTUAL SYSTEM after data input of the object under investigation and statement the tasks with the help of formal procedures will generate CORRESPONDING LOCAL INTELLECTUAL SYSTEMS able to solve the tasks better comparing to experts in the field specified.

Potential users of UNIVERSAL INTELLECTUAL SYSTEM can be Research and Development Centers involved in researches of the objects having characteristics of "large-scale" systems. (By Prigozhin, the result of a research is to be the table of experimental data and the research in itself is considered to be scientific when it results in creation of informative mathematical model and study of the object under investigation with the help of this model.

Potential users of LOCAL INTELLECTUAL SYSTEMS will be production plants, various enterprises, works and factories, medical centers which use or are planning to use Expert Systems in future.

7. Stage of development of the project.

Methodology of scientific research of the object which have different physical nature with use of experimental data obtained while observing ( "passive" experiment) has been developed and for many years tested with success in solving a number of various research and applied problems .

The package of computer programs for UNIVERSAL INTELLECTUAL SYSTEMS FOR INVESTIGATION OF COMPLEX OBJECTS is available allowing the following:

- development of LOCAL INTELLECTUAL SYSTEMS in any subject field specified;

- creation of informative mathematical objects, systems or phenomena of various physical nature using ether table of experimental data each line of which contains input and output variables;

- solving of acute scientific, research and applied tasks with the help of the models obtained.

LOCAL INTELLECTUAL SYSTEMS are available for differential diagnostics "gastric ulcer - gastric cancer", "hypertensia - renal hypertensia", etc. Mathematical models for them have been developed in cooperation with Military Medical Academy (St. Petersburg, Russia)

The following steps are needed:

- development of generalized structure of UNIVERSAL INTELLECTUAL SYSTEM which after input of information and statement of task can be easily transferred into specific LOCAL INTELLECTUAL SYSTEM;

- development of software for solving general-purpose tasks (algorithms for a number of tasks have been already developed) and integration of its structure with UNIVERSAL INTELLECTUAL SYSTEM.

- patenting of UNIVERSAL INTELLECTUAL SYSTEM.

8. Offer infringement.

Method of mosaic portrait, the base of intellectual system, is not asserted; the information of algorithm package providing completeness of the model with polynomial dependence of computer time from dimension have not been published; computer programs for creation of mosaic model are kept by the author.

9. Offer ownership belongs to Mark D. Kats

Information supplied by the Author December 1998.  Page last updated: January 19, 2004