The following is the list of tentative speakers for the seminar for Fall 2021.
September 1, 3:30 - 4:30, by Zoom
Speaker: Noah Giansiracusa (Bentley University)
Title: The Machine Learning Landscape
Abstract: I'll give a gentle tour of machine learning by explaining some of the main concepts and areas within the field as well as highlighting some exciting applications, and I'll leave ample time for students questions so that I can help guide you to resources and topics that you might delve into deeper later in the semester.
September 8, 3:30 - 4:30, by Zoom
Speaker: Seung Wook Kim (The University of Iowa)
Title: Application of machine learning algorithm to business research
Abstract: Machine learning algorithms are a very powerful tool that can uncover robust patterns in quantitative data. Researchers can use these meaningful patterns in order to formulate better hypotheses that are grounded in data. Indeed, machine learning models have been successfully used in many business research areas, including finance, management science, and marketing. During this talk, I will first talk about general concepts about using machine learning techniques and, by extension, show you how these powerful methods can be used for business research.
September 15, 3:30 - 4:30, by Zoom
Speaker: Seung Wook Kim (The University of Iowa)
Title: Using interpretable machine learning to understand consumer credit score
Abstract: Even though machine learning is a very powerful technique for analyzing big data and capturing complex patterns from the data, it is restricted when it comes to fully interpret the results of the models. The most accurate models for big data tend to be black-box models, which have long been criticized for their lack of interpretability. During this talk, I will discuss this interpretability issue and introduce an interpretable machine learning model that can solve this problem. Also, as a good example of an interpretable machine learning model, I will talk about my research project, which attempts to understand consumer credit scores.
September 22, 3:30 - 4:30, by Zoom
Speaker: Phillip Wesolek (Zendesk)
Title: Data Science in Industry
Abstract: Data Science is an exciting and challenging career path, which requires not only skills in mathematics, statistics, and computer science but also a keen understanding of business. In this talk, we will discuss some of the exciting challenges a Data Scientist faces. In particular, we will explore designing business-relevant metrics, training useful machine-learning models, and applying statistical models. Time permitting, we may discuss the role of software development in Data Science work.
October 6, 3:30 - 4:30, by Zoom
Speaker: Eric Brattain-Morrin (Butterfly Network)
Title: Computer Vision and Deep Learning
Abstract: Here’s an example of a function: given an image of lungs, output whether or not the patient has had COVID pneumonia. In your math classes, you are used to functions having clear numerical definitions, but in the messy real world, we have to find approximations, and it can be vital to know how accurate they are. Machine learning allows us to find approximations to functions that we don’t know how to numerically define. Seeing is believing, but how do we know what we are looking at? How can you write a program that takes an image labels each relevant thing? We will give a brief introduction to deep learning and how it is applied in computer vision with an emphasis on medical applications.
October 20, 3:30 - 4:30, by WebEx
Speaker: Vincent Martinez (CUNY, Hunter College)
Title: Parameter estimation for nonlinear dynamical systems
Abstract: An inherent problem in the modeling of natural phenomena is in obtaining accurate estimation of the parameters in the system. For instance, when studying fluid motion, the Navier-Stokes equations provides a model for a viscous, incompressible fluid flow in the form of a partial differential equation for the velocity of the fluid. The material parameter in this system is the fluid’s kinematic viscosity. Typically, the value of this parameter is determined empirically by experiments or statistically by data, and its exact value depends on the particular fluid itself. This poses the following fundamental mathematical question: Is it possible to recover the true value of the viscosity by having only partial information about the motion of the velocity itself? In other words, under what scenarios is it mathematically possible for one to recover this unknown viscosity. In this talk, we discuss a dynamic algorithm that allows one to learn the true values of parameters in certain nonlinear dynamical systems as partial observations are made on the system.
November 17, 3:30 - 4:30, by Zoom
Speaker: Taedong Yun (Google)
Title: Machine learning for biology and medicine, with a focus on genomics
Abstract: Research in natural science, such as physics and neuroscience, had a deep influence on the development of modern machine learning (ML) and artificial intelligence (AI) techniques. More recently, ML and AI have started "giving back", pioneering new methods to answer complex questions in science. In this talk I will showcase some examples of applying ML & AI to problems in biology and medicine, focussing on genomics, an interdisciplinary field studying the fundamental building blocks of life. I will start with a brief introduction to genomics, followed by a discussion of how state-of-the-art techniques in ML can be applied to genome research. I will also mention a few other astonishing advances in natural science made possible by ML and share general thoughts on ML & AI research. No background knowledge in biology or medicine is required.