February 24, 2016
Speaker: Dr. Nataliya Portman, Faculty of Science, Ontario Tech University
Title: The Art and Power of Data-Driven Modeling: Statistical and Machine Learning Approaches
Abstract: Ever-increasing volumes of data and computing power nowadays have led to the rise of machine learning, pattern recognition and data visualization methods. An idea of learning a mapping between predictor and outcome variables hidden in a high-dimensional training dataset has found applications in a variety of fields - image processing, medical imaging, bioinformatics, financial risk management, materials science to name a few. While machine learning seems to be put on a pedestal in data analysis (thanks to artificial intelligence enthusiasts), statistical learning still remains prevalent in certain areas such as clinical applications and epidemiology, for example. The choice of either statistical or machine learning approach depends on the example (training) dataset used. In this talk, I will discuss the differences between the two frameworks and illustrate their use on the assortment of problems drawn from my research experience in computational neuroscience and chemometrics. I will then give an overview of machine learning methods applied to the development of density functionals and materials property prediction based on data from quantum mechanical computations.
Title: The Art and Power of Data-Driven Modeling: Statistical and Machine Learning Approaches
Abstract: Ever-increasing volumes of data and computing power nowadays have led to the rise of machine learning, pattern recognition and data visualization methods. An idea of learning a mapping between predictor and outcome variables hidden in a high-dimensional training dataset has found applications in a variety of fields - image processing, medical imaging, bioinformatics, financial risk management, materials science to name a few. While machine learning seems to be put on a pedestal in data analysis (thanks to artificial intelligence enthusiasts), statistical learning still remains prevalent in certain areas such as clinical applications and epidemiology, for example. The choice of either statistical or machine learning approach depends on the example (training) dataset used. In this talk, I will discuss the differences between the two frameworks and illustrate their use on the assortment of problems drawn from my research experience in computational neuroscience and chemometrics. I will then give an overview of machine learning methods applied to the development of density functionals and materials property prediction based on data from quantum mechanical computations.
