Thursday, October 17 at 4:00pm
A. Paul Schaap Science Center, 1000
35 East 12th Street, Holland, MI 49423-3605
“Modeling Data with Machine Learning” by Dr. Darin Stephenson, Math Department
The prevalence of large sets of data in our technological society has given prominence to the issues involved in processing, displaying, modeling, and making decisions from data. The subject of data science lies at the interface of computer science, statistics, and mathematics, and has applications in almost every field of study. One particular branch of data science involves "machine learning", which is broadly defined as the process of programming computers to build predictive data models in an automatic way. Thus, a machine can "learn" a data model from a broad modeling framework by consideration of the available data, rather than having model parameters specified by a human. Often, such models have many thousands (or millions) of available parameters, and computers can sift through huge quantities of data in order to "train" model parameters in an incremental way. The availability of fast parallel computing (via GPUs or related cloud computing) often makes such models trainable in a reasonable amount of time. The goal is a model that both describes known training data well and also is effective in prediction for further data on which the model was not trained.
This talk will survey a few of the problems machine learning can address and give insight into some basic machine learning procedures. The talk will also highlight the 2-credit Math 295 course, "Machine Learning with Python", which will be offered for the first time in the upcoming spring semester.