A Whirlwind Tour of Unsupervised Machine Learning Methods with Cameron Fen

This is an Ann Arbor Computer Society presentation.
Sponsored by Arbormoon Software, Inc.:

Machine learning refers to the science of building mathematical models to predict elements of interest. The objective of this talk is to introduce these machine learning models to software engineers. I will focus on the lesser known, but more important problem of unsupervised learning. Unsupervised learning refers to learning interesting statistics using only x data without any labels y telling the model what the correct prediction at training time is. For example, the internet mostly contains unsupervised data suggesting that if we want to build artificial intelligence that can utilize the whole gamut of human experience, we need to have powerful unsupervised models. This presentation will assume only a basic understanding of machine learning up to a basic understanding of linear regression. However, I will still cover the cutting-edge models in machine learning including GANs, Normalizing Flows, Contrastive Models, and Transformers.

About the Presenter:

Cameron Fen is a data scientist, quantitative trader, and PhD student in economics. As a graduate student, he specializes in using the tools of deep learning to improve macroeconomic models. He has written papers on recurrent neural network economic forecasting, improvements in external validity using pooled data, and simulation-based inference for Bayesian estimation of dynamic macroeconomic models. He has presented at the Econometric Society, EcoMod, and Allied Social Science Association. Additionally, he was cofounder and head of research at AICM a quantitative trading startup which received support and funding from Mass Challenge, Nvidia, and the NexCubed Incubator. He has also been interviewed by Information Week, Search Enterprise AI, and Authority Magazine. You can find more about his research at cameronfen.github.io.