What your Machine Learning Books don’t tell you!

Abstract: Applications of machine learning methods have been celebrating a number of success stories in recent years. One may assume that building a power artificial intelligence is simple: Provide a data set to some (black box) algorithm from a book or library and get terrific outcomes. However, it is not that simple, as machine learning is not a silver bullet. This talk will cover a number of topics in machine learning that are usually not addressed in mainstream machine learning literature, however, they are relevant to both the academic and industrial worlds. These topics include systematic biases in data sets, choosing the right metrics to properly assess models, dimensionality reduction methods and comparability of models. We will analyze the challenges and discuss practical solutions. Furthermore, we will look at the technological singularity, the point in time when machines will become more intelligent than humans. We will discuss when or if the singularity can realistically be reached and what its consequences on society would be.

Speaker: Patrick GLAUNER is a PhD Student at the University of Luxembourg working on the detection of electricity theft in emerging markets through Machine Learning. His research was featured in New Scientist and cited in the McKinsey Global Institute discussion paper "Artificial intelligence: The next digital frontier?". He also holds two adjunct faculty appointments at the Universities of Applied Sciences in Karlsruhe and Trier. In parallel, he is pursing an MBA with Smartly. He graduated as valedictorian from Karlsruhe University of Applied Sciences with a BSc in Computer Science and obtained his MSc in Machine Learning from Imperial College London. He was a CERN Fellow, worked at SAP and is an alumnus of the German National Academic Foundation (Studienstiftung des deutschen Volkes). His current interests include anomaly detection, computer vision, deep learning, natural language processing, power supply, smart grid and time series.