Problem Solving with Multiple Criteria ? A New and Innovative Tool in ComputingSpeaker: Kalyanmoy Deb – MI, United States
Topic(s): Artificial Intelligence, Machine Learning, Computer Vision, Natural language processing , Software Engineering and Programming , Computational Theory, Algorithms and Mathematics , Information Systems, Search, Information Retrieval, Database Systems, Data Mining, Data Science , Applied Computing
Most practical search and optimization related problem-solving tasks involve multiple conflicting criteria, which all must be considered simultaneously during an optimization algorithm. A multi-criterion optimization problem gives rise to a set of Pareto-optimal solutions, which must first be found by a suitable algorithm and then analyzed to choose a single preferred solution. However, most common computational algorithms scalarize multiple criteria into a single criterion using a parameterization and employ a single-objective optimization algorithm to find its optimal solution. To generate a set of Pareto-optimal solutions, these scalarized single-criterion algorithms must be applied many times by changing the parameters of the scalarization, thereby proposing a computationally inefficient procedure. With the development of population-based evolutionary optimization methods, multi-criterion problems are better solved using a single application in finding multiple Pareto-optimal solutions simultaneously in an implicit parallel manner. In the distinguished lecture program, I plan to introduce the principles of multi-criterion optimization and present a number of popularly-used evolutionary multi-criterion optimization (EMO) algorithms in detail. Test problems, performance metrics and visualization methods commonly used in EMO studies will be introduced. The advantage of using EMO algorithms in practical problems will be demonstrated by showing results on a number of practical problems from industries (design and manufacturing) and societies (land use management). The latter part of the lecture will concentrate on advanced research topics in multi-criterion optimization including surrogate-assisted optimization, constrained optimization, bilevel optimization, robust and reliability-based optimization. Recent applications of EMO methods in machine learning problems, such as neural architecture search, adversarial DNN, etc. and machine learning methods in improving EMO’s performance will also be highlighted.
The participants will get comprehensive insights to the multi-criterion search and optimization field, its application domain and potential, public-domain codes, and related research ideas from the lecture. At different lectures, slight modifications will be made to suit the interests of the audience and to also cover various aspects of this emerging field.
About this LectureNumber of Slides: 50
Duration: 50 minutes
Languages Available: English
Request this Lecture
To request this particular lecture, please complete this online form.
Request a Tour
To request a tour with this speaker, please complete this online form.
All requests will be sent to ACM headquarters for review.