Program Synthesis and Transformation Techniques for Simulation, Optimization and Constraint.pdfVIP

Program Synthesis and Transformation Techniques for Simulation Optimization and Constraint Satisfaction Research Statement Thomas Ellman Vassar College Rutgers University Intro duction Computational scientists and design engineers face recurring problems of sp ecifying constructing testing and mo difying numerical programs Scientists often do not know in advance what ob jects and pro cesses must b e mo deled in numerical simulation programs Engineers often do not know in advance the b est way to formulate ob jectives constraints and search spaces that are used in pro grams for automatic design optimization Both typ es of professional typically take an exp erimental approach to the problem They develop one program exp erimentally examine its b ehavior and sub sequently mo dify it many times b efore arriving at a program that is adequate to the task at hand The pro cess of co ding and revising such programs is extremely timeconsuming and errorprone b ecause they are almost always written in conventional programming languages Furthermore the resulting programs are dicult to mo dify and reuse on other problems In practice this limits the range of problems that scientists and engineers are able to solve using computational techniques My research lies in a discipline known as Automated Software Engineering This is a eld of research that develops and applies Articial Intelligence and Formal Metho ds to problems of Software Engineering In my own research I fo cus on programs that involve simulation optimization andor constraint satisfaction I develop metho ds of synthesizing such programs from declarative sp ecication