Location
CS 2311
Event Description
Title: Multi-Modal Machine Learning: Text, Time Series and Beyond
 
A brief outline of the talk:
This talk will cover some state-of-the-art approaches in text mining, time series analysis and multi-modal learning, with hands on experiences on two concrete examples, classifying subtypes of lymphoma patients and predicting mortality of ICU patients. In the first part of this talk, I will show how to automatically convert both narrative text and time series into a unified graph representation and how to collect important subgraphs that are of interest. In the second part, I will show a multi-modal machine learning algorithm – subgraph augmented non-negative tensor factorization (SANTF), which is able to integrate both subgraph features and atomic features extracted from text and time series into the learning objective. These algorithms are applied to the above two experiments, and they demonstrate improved performance and better interpretability of the models. 
 

Short Bio:

Yuan Luo is a PhD Candidate at the Computer Science and Artificial Intelligence Lab at Massachusetts Institute of Technology. His research interests include machine learning, natural language processing, time series analysis, computational genomics and big data analytics, with a focus on medical and clinical applications. One of his current research projects is on tensor modeling of narrative text, time series and the Omic data. He proposed a framework for building clinical models that improves both accuracy and interpretability, by turning text and time series into graph representation and applying tensor factorization to mining graph features. This work was awarded the first prize at NLP Doctoral Consortium in 2013 Annual Symposium of American Medical Informatics Association. He is also interested in understanding the relationship among the federal funding, the topic evolvement in biomedical literature, and the biotechnology and pharmaceuticals corporate development. More projects and information can be found at http://people.csail.mit.edu/yluo/