Back to Pre-Meeting Webinars

Speaker: Clay Morrison

Presentation Title: AutoMATES: Automating Model Assembly from Text, Equations and Software

This Webinar is posted on the MSM YouTube Channel

Abstract
The Automating Scientific Knowledge Extraction (ASKE) DARPA program, led by Joshua Elliot, is developing a vision of future computational-driven science, in which computers understand scientific concepts and scientists develop, communicate, and evaluate their scientific models in a universal modeling framework. In the AutoMATES project, part of the ASKE program, is developing a framework for automating the extraction of scientific models from computer source code and linking them with code comments and machine reading of text and equations from scientific publications and technical documents. The framework extracts an integrated representation that enables quantitative comparison of models from different sources, automated model sensitivity analysis, and augmenting and composing model components into new model configurations.

Biosketch
Dr. Morrison's research is in machine learning and artificial intelligence, where he designs computer algorithms that learn rich, structured representations from unstructured data, such as images, video and text. These representations can then be used in complex decision making. Through the support of NSF, DARPA, AFOSR, and ONR, Dr. Morrison has led research projects that have developed machine learning algorithms, implemented and evaluated computer systems, and conducted empirical studies with human subjects. These projects have included developing machine reading technology to read scientific literature and computer source code to. extract and assemble computational models, computer systems that can be taught through natural human instruction, machine learning algorithms for learning structured, latent representations of data, and modeling the relationship between human facial expressions, emotion, and decision-making.

This is part of a series of recordings for the ML-MSM Meeting, see complete list of Pre-Meeting Webinars