Tom Hartvigsen

Logo

Optimizing for minimal expected negative feedback.

View My GitHub Profile

Previous research

Worcester Polytechnic Institute: 2016 - Present

More details to come for recent projects…

2018-2019

Time-aware attention

Attributed hierarchical attention

Early Classification

Disentangled recurrent memory curation

Mentoring one new PhD student on time series missing value imputation

Internship with UMass Medical School

2017-2018

Multimodal data fusion for MRSA detection

Joke Generation

Early time series classification

Led NSF REU project - Classifying time series with missing values

2016-2017

Leading NSF REU project - Early Prediction of MRSA Infections Using EHRs

Leading NSF REU project - Detecting Clostridium Difficile Infections Using Recurrent Neural Networks

Clostridium Difficile Risk Estimation (CREST)

Clinical data offers many opportunities for impactful and technically fascinating research. Beginning in August, 2016, I worked with Cansu Sen using the MIMIC III Intensive Care Unit Database. This database is publicly available, consists of 12 years of clinical data from ~58,000 admissions. You can request access to the database here. We focused on the detection of one infection, Clostridium Difficile, and therefore extracted patients who got this infection during their stay. As is required for supervised-learning algorithms, we also extracted a set of patients who did not get this infection and trained Logistic Regression, Random Forest, and Support Vector Machine models to detect patterns that indicate C. Diff. based on the following data sources:

We trained classifiers on each source of data, then merged their outputs to create a well-informed meta-learner to detect C. Diff. far in advance of the date of infection-confirmation. This work was published at the European Conference of Machine Learning (ECML), presented by Cansu Sen on September 19, 2017 in Skopje, Macedonia. Our paper can be found here.

University of Arizona REU: Summer 2015

Image segmentation to understand the phenological development of desert shrubs through drought periods

From June-August 2015, I took part in an NSF-funded Research Experience for Undergraduates at the University of Arizona in Tucson, AZ. I was stationed at BioSphere 2 and worked with Dr. Shirley Papuga in the School of Natural Resources and the Environment.

SUNY Geneseo: 2012 - 2016

2016: Scraped Movie Networks from IMDB

2015: Text Mining of Song Lyrics

2013-2014: Modeling Vaccination Strategies on Graphs