INTELLIGENT SIMULATION OF COMPLEX SYSTEMS USING IMMUCOMPUTING Svetlana P. Sokolova Ludmilla A. Sokolova St. Petersburg Institute for Informatics and Automation of RAS Kazan’, of February, 2008
WSC-6, of February, /24 Contents IMMUNOCOMPUTING POSSIBILITIES INDEX FORMAL IMMUNE NETWORK MATHEMATICAL BASIS APPLICATIONS OF THE IMMUNOCOMPUTING APPROACH
WSC-6, of February, /24 IMMUNOCOMPUTING POSSIBILITIES Immunocomputing represents a bridge between immunology and computer engineering, demonstrating how quantitative advancements in immunology can form the basis for a new computing paradigm Immunocomputing possibilities: capacity for memory the ability to learn and recognize, and make decisions in conditions of uncertainty and incomplete information an excellent information-processing model for designing a powerful computing system
WSC-6, of February, /24 RISK INDEX Indices reduce large quantities of variable data (uncertainty, multidimensional and so on) relating to a complex dynamic systems into a single value (Data Fusion) to achieve a solution to a practical problem Sometimes this is the only way to represent a system and predict risks and trends Risk Index is overall index indicating an irregular situation
WSC-6, of February, /24 FORMAL IMMUNE NETWORK STRUCTURE Index values given for training sample Training module «antibody» «antigen» Pattern recognition module «binding energy» Module of index coefficients optimization Calculating of index values Learning sample “immunization” Index coefficients
WSC-6, of February, /24 TRAINING MODULE Forming of training matrix Singular value decomposition Selection of vector for indices formation Selection of the most significant indices
WSC-6, of February, /24 MODULE OF INDEX COEFFICIENTS OPTIMIZATION Forming of index coefficients matrix Calculating of pseudo inverse matrix Calculating of optimal index coefficients Singular Value Decomposition
WSC-6, of February, /24 Basic Algorithm of Immunocomputing (in pseudo code) Learning // data mapping into FIN space { to receive a learning sample to form learning matrix to calculate SVD of the learning matrix //SVD–singular value decomposition// } Recognition // data classification in FIN { to receive a situation vector //pattern to map a vector in FIN space to find the closest FIN point to assign a vector the closest FIN point class }
WSC-6, of February, /24 Monitoring Plague problem
WSC-6, of February, /24 Plague Risk Index
WSC-6, of February, /24 The analysis of credit status of the borrower - its ability to pay off under the promissory notes completely and in time CREDIT RISK INDEX 1. Interval data: group1 client group2 2. Multidimensional data:
WSC-6, of February, /24 APPLICATION OF THE CREDIT RISK INDEX IndicatorEvaluation 1.Checking Account with Bank1 balance less or equal to zero 2 balance between DM 0 and DM more than DM 199 or used more than one year for regular salary payments 4 no such account with bank 2.Duration of Loan1 <= <... <= <... <= <... <= <... <= <... <= <... <= <... <= <... <= <... <= Foreign Worker1 Yes 2 No
WSC-6, of February, /24 RESULT
WSC-6, of February, /24 Conclusion Application of Immunocomputing approach significantly increases a potential of realization in real systems The considered technologies implementation to a broader problems class, including monitoring systems of various size and orientation
WSC-6, of February, /24 Thank for your attention