dc.contributor.author | Misener, Ruth | |
dc.date.accessioned | 2018-10-12T20:56:32Z | |
dc.date.available | 2018-10-12T20:56:32Z | |
dc.date.issued | 2018-09-26 | |
dc.identifier.uri | http://hdl.handle.net/1853/60481 | |
dc.description | The Suzanne C. and Duncan A. Mellichamp Distinguished Lecture was presented on September 26, 2018 from 3:00 p.m.- 4:00 p.m. in the Molecular Science and Engineering Building (MoSE), Room G011, Georgia Tech. | en_US |
dc.description | A gift from Suzanne C. and Duncan A. Mellichamp established this lecture series at Georgia Tech’s School of Chemical & Biomolecular Engineering in 2016. Duncan Mellichamp, who earned his bachelor’s degree at Georgia Tech in 1959 and PhD from Purdue University, became a research engineer for DuPont before being recruited to help create the chemical engineering department at the University of California-Santa Barbara. Author of more than 100 research publications, Professor Emeritus Mellichamp mentored more than 50 graduate students to degrees. The Mellichamps have endowed scholarships at Georgia Tech in the School of Chemical & Biomolecular Engineering and the School of Materials Science and Engineering. These scholarships benefit undergraduate students who demonstrate scholastic excellence. | en_US |
dc.description | Dr. Ruth Misener is a Senior Lecturer (U.S. equivalent: Assistant/Associate Professor) in the Computational Optimisation Group of the Imperial College London Department of Computing. Misener currently holds an EPSRC Early Career Research Fellowship (2017–22) and previously held a Royal Academy of Engineering Research Fellowship (2012–17). Her research awards include the W. David Smith, Jr. Graduate Student Paper Award (2014), the Sir George Macfarlane Medal as the overall winner of the 2017 RAEng Engineers Trust Young Engineer of the Year competition, and the 2017 AIChE 35 Under 35 Innovation category. Misener serves as a Director of the AIChE Computers & Systems Technology Division (2016–18) and on the editorial boards of Computers & Chemical Engineering, Optimization & Engineering, and Mathematical Programming B. | en_US |
dc.description | Runtime: 56:43 minutes | en_US |
dc.description.abstract | Surrogate models are widely appreciated in chemical engineering. The typical setting focuses on
expensive-to-evaluate, possibly uncertain functions. Resources are typically limited, so effective
decision-making requires data-efficient learning. The data science and statistical machine learning communities typically focus on models learned solely from observed data. But chemical engineering applications may also require explicit, parametric models, e.g. modeling known process constraints, operations constraints, and cost objectives. So work has integrated semi-algebraic functions with those learned from data or developed semi-physical modeling techniques. We consider three new probabilistic modeling applications and extend methodologies to meet these hybrid situations: Design of experiments for model discrimination. We bridge the gap between classical, analytical methods and Monte Carlo-based approaches. Classical methods may have difficulty managing nonanalytical model functions and data-driven Monte Carlo approaches come at a high computational cost. We replace the original, parametric models with probabilistic, non-parametric Gaussian process surrogates learned from model evaluations. The surrogates are flexible regression tools that extend classical analytical results to non-analytical models, while providing us with model prediction confidence bounds and avoiding the computational complexity of Monte-Carlo approaches. Multi-objective optimization. We make novel extensions to Bayesian multi-objective optimization in the case of one analytical objective function and one black-box, i.e. simulation-based, objective function. The resulting method has been applied to a bone neotissue application and a more general test suite. Scheduling plant operations under uncertainty. For processes with equipment degradation, we use Gaussian processes to approximate large-scale, mixed-integer optimization problems. We close by offering a broad outlook on applying probabilistic surrogate models to chemical engineering. | en_US |
dc.format.extent | 56:43 minutes | |
dc.language.iso | en_US | en_US |
dc.publisher | Georgia Institute of Technology | en_US |
dc.relation.ispartofseries | School of Chemical and Biomolecular Engineering Seminar Series | en_US |
dc.relation.ispartofseries | Mellichamp Distinguished Lecture | en_US |
dc.relation.ispartofseries | School of Chemical and Biomolecular Engineering Seminar Series | |
dc.subject | Model selection | en_US |
dc.subject | Multi-objective optimization | en_US |
dc.subject | Probabilistic modeling | en_US |
dc.title | Gaussian Processes for Hybridizing Analytical & Data-Driven Decision-Making | en_US |
dc.title.alternative | Gaussian Processes for Hybridizing Analytical and Data-Driven Decision-Making | en_US |
dc.type | Moving Image | |
dc.contributor.corporatename | Georgia Institute of Technology. School of Chemical and Biomolecular Engineering | en_US |
dc.contributor.corporatename | Imperial College, London. Department of Computing | en_US |
dc.type.genre | Lecture | |