Pythia: a Knowledge Based System for Intelligent Scientiic Computing
نویسندگان
چکیده
1. ABSTRACT Domain speciic Problem Solving Environments (PSEs) are the key new ingredients that will aid in the widespread use of Computational Science & Engineering (CS&E) systems. Each PSE consists of a well deened library that supports the numerical and symbolic solution of certain mathematical model(s) characterizing a speciic discipline, together with an easy to use software environment. This environment should ideally interact with the user in a language \natural" to the associated discipline, and provide a high level abstraction of the underlying, computationally complex, model. However, it appears that almost all extant PSEs assume that the user is familiar with the speciic functionality/applicability of the PSE. Their primary design objective is to support some form of high level programming with predeened state-of-the-art algorithmic infrastructure. As the functionality of these systems increases, the user is expected to make complex decisions in the paramet-ric space of the algorithmic infrastructure supported by the PSE. In this paper we describe a knowledge based system, PYTHIA, to automate this decision making process and aid in providing a high level abstraction to the user. Speciically, PYTHIA addresses the problem of (parameter, algorithm) pair selection within a scientiic computing domain assuming some minimum user speciied computational objectives and some characteristics of the given problem. PYTHIA's framework and methodology is general and applicable to any class of scientiic problems and
منابع مشابه
Neural and Neuro - Fuzzy Approaches to Support \ Intelligent " Scienti c Problem Solving
Scientiic computing uses computers, especially High Performance Computing (HPC) systems, to solve complex mathematical equations which model physical phenomena. Using these systems now requires expert knowledge in a variety of elds of computer science, such as parallel computing and numerical methods. This often makes application scientists, who have the domain expertise to devise the mathemati...
متن کاملNeuro - Fuzzy Systems for Intelligent Scienti c Computation
Intelligence has been envisioned as a key component of future problem solving environments for sci-entiic computing. This paper describes a computationally intelligent approach to address a major problem in scientiic computation, i.e., the eecient solution of partial diierential equations (PDEs). This approach is implemented in PYTHIA-a system that supports smart parallel PDE solvers. PYTHIA pr...
متن کاملPYTHIA : A Knowledge Based System to
Problem Solving Environments (PSEs) interact with the user in a language \natu-ral" to the associated discipline, and provide a high level abstraction of the underlying , computationally complex, model. The knowledge based system, PYTHIA, addresses the problem of (parameter, algorithm) pair selection within a scientiic computing domain assuming some minimum user speciied computational objective...
متن کاملFull Day Tutorial T 3 : RELATIONAL SEMIOTIC METHODS FOR DE - SIGN OF INTELLIGENT SYSTEMSL
material presented at the tutorial as well as a useful list of references. Two-valued Boolean (Crisp) as well as their many-valued extensions (fuzzy relations) play increasingly important role in representation and computing with engineering and scientiic data and knowledge. Relations provide important tools in dealing with modal and belief logics used in AI. In intelligent systems and Knowledg...
متن کامل