Christopher Gotwalt, PhD, SAS Institute | Philip Ramsey, PhD, Predictum Inc. | Wayne Levin, Predictum Inc.
Self-Validating Ensemble Modeling (S-VEM) is an exciting, new method that delivers machine learning accuracy to Design of Experiments (DOE) and has many applications in manufacturing and chemical processes.
Machine learning methods are valued because they produce excellent predictive models, but until now they have been disqualified from use with small data sets, including designed experiments, because of the limited amount of data. The limitations have been:
· DoEs could not be validated, so they were less reliable.
· The relationship between the models and the designs were essentially fixed. Expanding the terms in the model and dealing with noisy data required running additional experiments.
· Typically experiments could only accommodate up to second-order parameters, yet the relationships among parameters is typically higher.
S-VEM generates robust, higher-order models that yield greater insights and characterization using considerably fewer runs, thereby saving time and money and reducing risk.
In this talk, we’ll provide an overview of S-VEM as a new, advanced machine learning method, demonstrate a couple of use cases using our new S-VEM analytical product, and suggest how you can explore this method further.
Who Should Attend:
Engineers, scientists and their management who are engaged in research, problem solving and process development and characterization