Speaker: 

Katie Rainey

Institution: 

SPAWAR Navy Lab, San Diego

Time: 

Wednesday, February 27, 2019 - 5:00pm

Location: 

RH 306

Classification functions are central to most applications of machine learning. Recent advances in deep learning algorithms have shown impressive results on image recognition and other classification tasks, motivating researchers to apply such algorithms to all manner of problem domains. However, many practitioners fail to understand how the algorithms work, and in particular how they operate on a given dataset. In this talk I will discuss deep learning classifiers from the perspective of a mathematical function, to shed light on how the algorithms operate on data at a fundamental level. I will also discuss several on-going research efforts demonstrating techniques that enable practitioners to gain insights into the complexity and behavior of their data and algorithms. With luck, the talk will stimulate discussion about the understanding — of the task, the data, and the algorithm — necessary to confidently deploy deep learning classifiers in operational and safety-critical scenarios.