- Week 1: Jan 07th: Pratool Bharti, Assistant Professor, Department of Computer Science, Northern Illinois University, DeKalb, IL, USA
- Week 2: Jan 14th: David Koop, Assistant Professor, Department of Computer Science, Northern Illinois University, DeKalb, IL, USA
- Week 3: Jan 21st: Robert Tell, Vice President, Bioinformatics, Tempus Labs, Inc., Chiacgo, IL, USA
- Week 4: Jan 28th: Mark Potosnak, Professor and Department Chair, Department of Environmental Science and Studies, DePaul University, Chicago, IL, USA
- Week 5: Feb 04th: Benjamin Langmead, Associate Professor, Department of Computer Science, Johns Hopkins University, Baltimore, MD, USA
- Week 6: Feb 11th: Dan Knights, Associate Professor, Department of Computer Science and Engineering, University of Minnesota, Twin Cities, NM, USA
- Week 7: Feb 18th: TBD
- Week 8: Feb 25th: TBD
- Week 9: Mar 04th: Debzani Deb, Associate Professor, Department of Computer Science, Winston-Salem University, Winston-Salem, NC, USA
- Week 10: Mar 11th: TBD
- Week 11: Mar 18th: FINALS WEEK – NO SEMINAR – GOOD LUCK WITH FINALS!
TITLE: Context-aware Machine Learning Models for Personalized and Public Health
ABSTRACT: Innovations in designing Machine Learning models (and creating novel applications) have been growing at a rapid pace in the last decade. More recently, there is an earnest interest in context-aware learning, where the goal is to carefully extract limited contexts within the domain of interest during model development and execution. Such contexts can be broad and also domain specific – for example physiology of the human body in classifying physical activities; evolution of anatomies in classifying biological species; dynamically changing outdoor environments during modeling and optimization of flood detection system; and much more. The motivation for integrating contexts during learning not only improves accuracy, but also saves on implementation/ execution cost. However, this process is challenging and often requires multi-disciplinary expertise.
In this talk, I will highlight my recent technical contributions in this space. First, I will present my research on designing physiology-aware learning models to accurately classify complex human activities using wearable devices that are significant for personalized elder care. The innovation here is careful integration of multi-modal inertial sensory data from multiple wearable devices emplaced across multiple positions in the human body, and finally integrating human physiology into decision making. Second, I will present my results on neural network models to classify genus and species types of mosquitoes from smart- generated images taken by experts or by ordinary citizens. The innovation here lies in extracting contextually relevant anatomies (e.g., head, thorax, wings, legs) from mosquito images, and assigning appropriate weights to only the most critical anatomical component(s) for accurate classification. This work is expected to have significant impact in automating mosquito surveillance and related public health efforts in the US and across the globe. Towards the end, I will briefly explain my work on vision guided flood warning system to detect water level in the stream under dynamic environmental conditions.
TITLE: Reproducible Computational Notebooks
ABSTRACT: Computational notebooks provide a setting where users can rapidly examine and evaluate intermediate outputs as solutions are explored through blocks of code named cells. However, these explorations can be difficult to reuse or reproduce in the future, especially by someone other than the original author. We have investigated techniques both to discover provenance from saved notebooks themselves and to enforce greater structure in new notebook explorations. While notebooks encode clues about execution patterns and often share similar structures, we have found there is usually not enough information to infer the provenance of past notebooks. To address this issue, we have built dataflow notebooks as a way to clearly structure dependencies between cells. This makes computations more traceable and reproducible, but brings some interesting usability challenges. We propose new methods to display and refer to cell outputs in order to minimize rewriting while allowing evolution of the code.
TITLE: Liquid biopsy testing at Tempus and modalities of therapeutic intervention
ABSTRACT: This talk will review the variety of technical challenges associated with liquid biopsy testing in the field and a review of a recent publication from Tempus labs describing analytical solutions to these problems. Content will largely be from the publication, Validation of a liquid biopsy assay with molecular and clinical profiling of circulating tumor DNA
TITLE: Enhancing undergraduate data-science skills and participation through community partnerships
Abstract: The Metropolitan Chicago Data-science Corps (MCDC) is funded through the National Science Foundation to promote practical data-science abilities and increase participation from underrepresented groups. From a student perspective, there are three phases: a year-long sequence of introductory data-science pathway courses, a data-science practicum class and a summer internship with salary support. The latter two phases incorporate real-world data-science projects that are solicited from community partners. There are five Chicago-area universities involved: Northwestern (the lead), DePaul, Northeastern Illinois, Chicago State and University of Illinois’s ischool. Each university has its own pathway sequences and individual data-science practicum courses. Students from different participating universities will come together in teams led by faculty mentors to work with non-profit community partners. These partnerships will focus on social justice, environmental issues and community health. The vision is that students will be drawn towards understanding data science through real-world learning and working on problems that are inclusive and support the common good.