Quantifying Reproducible Machine Learning
Event date:  December 14, 2020 4:00 PM  to 6:00 PM

Abstract

Dr. Raff has performed a retroactive study of 255 attempts to implement published machine learning algorithms. In doing so he has developed a set of features to attempt to quantify factors that impact the likelihood of being able to reproduce a machine learning paper from scratch - without access to the original source code. Analyzing these results we can begin to build evidence for and against different opinions that have been argued about the nature of reproducibility. Beyond answering statistical questions about what has a significant relationship with being reproducible, the further be extended with careful ML modeling to start to change how we measure reproducibility: moving beyond the binary view that some work is or is not reproducible.   


About the Speaker

Hello, my name is Edward Raff. I'm a Chief Scientist at Booz Allen Hamilton, where I lead our Machine Learning research group. I'm also a Visiting Professor at UMBC were I work closely with the DREAM and IRAL labs, and teach courses in the Data Science department. I received my PhD in Computer Science from University of Maryland, Baltimore County (2018) and was supervised by Professor Charles Nicholas. I graduated from Purdue University with a Bachelors (2012) and Masters (2013) degree in Computer Science. Please see my linkedin for more of my work history.


About the Series

The Lecture Series is organized every winter term and consist of talks with international speakers. The speakers are experts from various fields and professions. The talks are aimed at covering the whole spectrum of visual computing and at discussing the relevance of quantification. This years lecture series will be held entirely online.

This is an online event only.

Registration: If you would like to attend, please email Leonel Merino to register.