Dr. Maxwell J. Robb, Assistant Professor of Chemistry
Dr. Marianne Bronner, Albert Billings Ruddock Professor of Biology
Development and Implementation of Captive Trajectories in the NOAH Water Tunnel Laboratory, Alex Wuschner
Abstract: The NOAH Water Tunnel Laboratory is currently used to improve our understanding of various aspects of turbulence and fluid mechanics. The goal of this project is to understand the capabilities of a newly installed technology and the opportunities it presents for enhancing the the use of the NOAH Laboratory in the future. The recently installed Captive Trajectory System (CTS) , a cyber-mechanical system capable of moving and rotating within the water tunnel, allows us to explore new methods of simulating objects moving in complex, turbulent fluids. By harnessing the ability to program how the CTS behaves, the system was shown effective in modeling the motion of an object in real time as variable forces were applied to it. Through simple, controlled examples, we discovered the ability of the CTS to accurately model various types of motion, such as that of a mass-spring-damper system, or a planet orbiting a sun. In more complex examples, the CTS was able to simulate the general behavior of an airfoil in the wake of a cylinder with vortex shedding. The examples explored over the course of the project have proven that the CTS can be used as a useful experimental tool and will open the door to new methods of studying turbulence and unsteady aerodynamics in the future.
Detection of Volume Changes in Greenland’s Marine Terminating Glaciers, Gurjot Kohli
Abstract: Increased melting of marine-terminating glaciers of the Greenland Ice Sheet could lead to unstable dynamic ice mass loss, further accelerating global sea level rise. To better understand the historical context of the present-day widespread glacier front retreat, we mapped frontal positions for two years, 1994 and 2017, for 146 and 251 major outlet glaciers, respectively. Front position locations were identified using optical remote sensing imagery from the Landsat 5 (1994) and Landsat 8 (2017) satellites. Of the fronts surveyed, we find that 86% retreated between 2017 and 1994 and 62% retreated between 2017 and 2015, when the most recent glacial front positions were recorded. In addition, ice surface elevation differences for four dynamically-thinning glaciers were calculated using data collected from NASA’s Oceans Melting Greenland (OMG) mission in 2016 and 2017 by the Glacier and Ice Surface Topography Interferometer (GLISTIN-A). In each case, ice surface lowering was observed in excess of 10 m within the 10km-wide swath of GLISTIN-A near the glacier front. This evaluation of glacier change advances our understanding of ocean-driven Greenland ice mass loss.
Eye Measurements as Objective Measures of Flow Experience, Salma Elnagar
Abstract: Flow experience is achieved when a person is said to be “in the zone” as they achieve a fit between skill-and-challenge level in a certain activity. Many of these activities, such as video games, music, or athletic competitions, involve the participation or company of other people. Thus, interpersonal communication could be related to reaching the mental state of flow. Previous research has linked brain areas like the inferior frontal gyrus to individual flow. However, no experiments have yet been conducted on interpersonal flow. The large-scale purpose of this project is to fill this gap by identifying the neural correlations of interpersonal flow. In order to do so, objective measurements of the inherently subjective flow experience must first be identified. In this study, we evaluated eye measurements such as pupil diameter, blink rate and eye fixations as objective measures for individual flow. Two versions of our experiment using a musical notes game were conducted on ten and six subjects respectively with eye measures being collected using Eyelink 1000. While there was a significant difference in gaze direction among various gaming levels, there appeared to be no significant change in blink rates or pupil size. This suggests that gaze direction could be an objective measurement of flow experience.
Machine Learning for Cybersecurity: Network-based Botnet Detection Using Time-Limited Flows, Stephanie Ding
Abstract: Botnets are collections of connected, malware-infected hosts that can be controlled by a remote attacker. They are one of the most prominent threats in cybersecurity, as they can be used for a wide variety of purposes including denial-of-service attacks, spam or bitcoin mining. We propose a two-stage, machine-learning based method for distinguishing between botnet and non-botnet network traffic, with the aim of reducing false positives by examining both network-centric and host-centric traffic characteristics. In the first stage, we examine network flow records generated over limited time intervals, which provide a concise but partial summary of the complete network traffic profile, and use supervised learning to classify flows as malicious or benign based on a set of extracted statistical features. In the second stage, we perform unsupervised clustering on internal hosts involved in previously identified malicious communications to determine which hosts are most likely to be botnet-infected. Using existing datasets, we demonstrate the feasibility of our method and implement a proof-of-concept, real-time detection system that aggregates the results of multiple classifiers to identify infected hosts.
Convolutional Neural Networks as Efficient Emulators for Atmospheric Models, Berlin Chen
Abstract: We used Convolutional Neural Networks (CNNs) to emulate the physics of the atmosphere in order to bypass solving partial-differential equations (PDEs) explicitly, which cuts down on computational cost. This is important because in the past, the models used to produce reliable weather forecasts required a computationally complex calibration.
We let the CNNs learn on a 4-dimensional (longitude x latitude x height x time) geophysical dataset, with a separate CNN for each height index. After a series of experiments we conducted following implementation, we found that zero-padding the data, varying time scale, and changing sample space had little effect on the CNNs’ performance, and that there was little correlation between training data size and error. We also observed that given the same training-data size, the CNNs with a more complex configuration (more sets of weights) actually performed worse.