CS majors Michael Altschuler (Class of 2019), Garrett Beatty (2019), and Ethan Kochis (2020), presented two research papers at the IEEE ICSC 2019 conference in Newport Beach, California, held January 29 – February 1. Both papers were written in collaboration with CS faculty member Dr. Michael Bloodgood as part of the students’ mentored research conducted in fall 2018. Both papers were supported by TCNJ’s Support of Scholarly Activities (SOSA) program and by use of the ELSA high performance computing cluster at TCNJ, supported by the National Science Foundation under grant number OAC-1828163. Dr. Bloodgood also attended the conference and served as a session chair during the conference.
Michael Altschuler and Dr. Bloodgood co-authored a paper titled: “Stopping Active Learning based on Predicted Change of F Measure for Text Classification.” In this paper, a new stopping method called Predicted Change of F Measure is introduced that provides users an estimate of how much performance of machine learning models can be expected to change at each iteration of learning.
Garrett Beatty, Ethan Kochis, and Dr. Bloodgood co-authored a paper titled: “The Use of Unlabeled Data versus Labeled Data for Stopping Active Learning for Text Classification.” This paper compares and contrasts the advantages and disadvantages of methods for stopping machine learning of text classification systems using three different information sources that have not been compared and contrasted before, with the perhaps surprising result that methods that use unlabeled data are more effective than methods that use labeled data. This paper was also supported by TCNJ’s Mentored Undergraduate Summer Experience (MUSE) program.
More information about IEEE ICSC 2019 can be found at: https://semanticcomputing.wixsite.com/icsc-2019