2020-02-08: My Impression to the 2020 ODU Undergraduate Research Symposium

Today, I went to the 2020 ODU Undergraduate Research Symposium (URS). This year, the symposium is held in the Learning Commons, Perry Library, running from 8 am to 12:30 am, including both poster and oral sessions. According to the key organizer Eddie Hill, there are 45 posters. The total number of exhibitions is 125, including posters, oral, and art exhibits. I sat in two oral sessions and went through almost all posters. There are several works that drew my attention, especially the ones that apply machine learning (ML) and neural networks (NN) on cross-disciplinary research. I briefly describe them below.

1. Radio Frequency Signal Classification and detection of Drones based on Machine Learning
    by Michael Nilsen, mentored by Dr. Sachin Shetty, Electrical and Computer Engineering
    The goal of the research is to develop a classifier to distinguish between UAS (unmanned aircraft system) signals and non-UAS signals in multiple bands. For a particular band (e.g., 5m), the work uses kNN for binary classification. The results indicate that the model accuracy depends on the band frequency and existence of shallow and obstruction.
    This looks like a good example of applied machine learning on electrical signal processing. One question that is not clear is what is the role of clustering here.

2. Validity and Reliability of Eye-tracking as a Measure of Impasse in Creativity and Problem Solving
     by Emily E. Russell, mentored by Ivan K. Ash, Cognitive Psychology

     The goal of this research is to analyze fixation as a pattern of impasse. Basically, they ask participants to try to solve several Insight Problems (Triangle problem or Katona Square) and measure how long they stare at each circle or stick. This is an on-going work (unfinished) and the poster is not well designed. However, the way they collect data is interesting. Their eye-tracking tools is a horizontal bar instead of glasses. The presenter told me that one drawback of the glasses is that they tend to slip up and down which may cause measuring inaccuracy. Dr. Sampath Jayarathna at the CS department can verify this.
      This student is supported by the ODU PURS program.

3. Particle Trajectory Classification and Prediction Using Machine Learning
    by Angelos Angelopoulos, Polykarpos Thomadakis, mentored by Gagik Gavalian and Nikos Chrisochoides

    The undergraduate presenter was mentored by a researcher at Jefferson Lab & ODU (Gagik) and a CS professor (Nikos). The research attempts to do 2 things: (1) predict particle trajectories based on previous trajectory and (2) classify trajectories as being valid or not. They used Gated Recurrent Unit (GRU) for Task (1). They applied 3 methods (Extremely Randomized Trees or ERT, Multilayer Perceptron, and a Convolutional Neural Network) to build a binary classifier for Task (2). For Task (2), the experimental results using MLP, and CNN are better than ERT by 3% using the A1 Metric, but the training accuracy of ERT is 5-6% higher than MLP and CNN. The time to predict per sample is 4-5 microseconds for MLP and ERT and 1.2 milliseconds for CNN. It looks like MLP is a very promising method for realtime prediction. The result for Task (1) is also promising with a Loss of 1.18 (1.18 sensors between the actual and predicted particle tracks).
    This seems a successful example of applying ML to physics.

4. Factors that Influence City Micromobility: An Investigation of the Use of Shared E-Scooters Among College Students
    by Qwe'Vontae Eure, mentored by Jing Chen, Psychology
 
    This is the only poster with NSF logo and I know Dr. Jing Chen has successful applications or NSF grants. The goal of this research is to investigate the motivation of using E-Scooters on campus and various factors that could impact the use of E-Scooters. The samples are only drawn from ODU.

5. Orofacial Manifestations of Lyme Disease: A Systematic Review
    by Kelsey Jones, mentored by Brenda Bradshaw

    This is an oral session. I have been trying to collaborate with Brenda and Holly Gaff to apply ML and NLP methods for mining scholarly articles in biomedical science papers, focusing on Lyme Disease. I knew Brenda was drafting a review paper on Lyme disease so this is a good opportunity to know the progress. The presenter presented orofacial manifestations reported in the surveyed articles. I asked a question on how many patients exhibiting the orofacial symptoms were misdiagnosed by doctors (either dentist or not)  and then later diagnosed as Lyme Disease. The presenter could not answer this question. Writing survey papers are non-trial and needs lots of work. It would be useful to develop a tool to automatically generate survey text by mining relevant papers.

Overall, I think the symposium is successful. The event was scheduled pretty early but the attendence was not bad. They have 10 judges evaluating posters. They will rate posters to 3 places and the presenters will receive monetary awards offered by Dean Metzger.

Jian Wu
https://www.cs.odu.edu/~jwu/