Jared Knofczynski A Multi-Task Weak Supervision Framework for Internet Measurements
The ability of machine learning (ML) systems to identify patterns in data is of growing importance to researchers in all fields, especially in the domain of Internet measurements. As our reliance on the Internet continues to grow, ML solutions to networking problems continue to be invaluable in ensuring the sustained performance of networked systems around the globe. One key issue network researchers face is a lack of labeled training data, particularly at scale. Traditional labeling strategies are less effective in this domain, as labeling network data often requires significant domain expertise that crowdsourced labeling resources do not possess, and the vast quantities of data make large-scale manual annotation infeasible. Additionally, many ML applications require multiple tasks to operate effectively, resulting in the multiplicative growth of training times as the number of tasks increases, but the lack of information sharing between tasks means that potentially useful information may be discarded if deemed irrelevant for the task at hand, when it could be useful to another model training on the same dataset. Given these challenges, we propose ARISE, a multi-task framework capable of leveraging weak supervision strategies in the form of labeling functions to label vast quantities of network data while sharing information between tasks to decrease training times, improve classification accuracy, and reduce the influence of hidden biases found within sets of training data.
Riley Monsrud Determining Physical Characteristics of the Asteroid 572 Rebekka through Analysis of its Lightcurve
Here we present observations of the asteroid 572 Rebekka that were obtained in August 2020 at Pine Mountain Observatory (PMO). The target was observed with the 0.35m Robbins telescope at PMO, using a Sloan g’ filter, for a total of 6 hours over two nights. The observations produced 436 images of the target which were then analyzed to produce a “lightcurve” of the asteroid. Through photometric analysis of the lightcurve, we have produced a 3-dimensional model of the asteroid which is presented here. Using the programs MPO Canopus (MPO) and Aperture Photometry Tool (APT), photometric estimates of the asteroid’s brightness over time are plotted in order to extract the rotation period as well as the shape of the target. To calibrate this data, we compare the asteroid to multiple stars of constant brightness within the same image. This process, known as “relative photometry”, allows us to remove atmospheric effects due to air quality, light pollution, and changing air mass. An estimate for the change in magnitude due to air mass, commonly known as the extinction coefficient, is also made. These findings give confidence in PMO’s ability to provide research-grade data and serves as an exercise in analyzing and reducing large sets of data. As a collaboration with Kobe University in Japan, this is a continuing project that looks to familiarize students with data analysis, calibration, and astronomical concepts.
Lucy Roberts Quantifying Glacial Melt and Movement Using Remote Sensing in Greenland's Sermilik Fjord
In Greenland’s fjords, large icebergs have been shown to be an indicator of oceanic circulation. However, previous reports published by Dr. Dave Sutherland’s Ocean and Ice group concluded that there is large variation of these flow systems seasonally and interannually. Providing reliable analysis of fjordic flow regimes in a specific fjord requires long-term data in order to mitigate annual fluctuations. Working along with Dr. Sutherland, I have been updating previously published research (2014 Article: “Quantifying flow regimes in a Greenland glacial fjord using iceberg drifters”) that used a very limited dataset.
I have been working with Dr. Sutherland to take these GPS data spanning 2012-2019. The previous publication used 10 GPS units from 2012 and 2013 to interpret iceberg motion in the context of mean fjord circulation. In aggregating eight years’ worth of data, we will be able to analyze information from >30 GPS devices to inform questions of recirculation and in-/out- fjord variations in velocity as they relate to flow variability, while mitigating noise from annual fluctuations. As large conduits of freshwater, the movement of icebergs and their interactions with the surrounding oceans are increasingly important when analyzing the impacts of global ocean warming. This project can provide the analysis needed to create and run more accurate models of fjord circulation and the ultimate fate of freshwater delivery from Greenland.
Nobuyuki Tamai Coordinated observations of asteroids by Pine Mountain and Nishi-Harima Astronomical Observatories
Coordinated broadband photometric measurements of the asteroid 665 Sabine were obtained in August 2020 from the Nishi-Harima Astronomical Observatory (NHAO) and Pine Mountain Observatory (PMO) using a 0.6 meter telescope (NHAO) and the 0.35m Robbins telescope at PMO. 665 Sabine has a rotational period of 4.294 hours, semi-major axis of 3.14 AU, and diameter of ~51km. In total, these observations produced 180 images from PMO, and 280 images from NHAO. These 460 images of the target were then analyzed to produce a “lightcurve” of the asteroid, where photometric estimates of the asteroid’s brightness over time are plotted in order to extract the rotation period as well as the shape of the target. At most locations, 665 Sabine cannot be tracked for more than two rotations. Continuous observation of rotating asteroids over several rotation cycles is necessary for determination of basic asteroid properties - such as the shape, surface properties, and rotation period. To obtain continuous coverage of an asteroid for more than two rotation cycles, multiple observing sites (separated by ~6 h in longitude) are needed. As a collaboration, NHAO and PMO work together to obtain data on asteroids that span several rotation periods. NHAO is operated by the University of Hyogo and located in Sayo, Japan (lat ~ 35 N, long ~ 134 E). PMO is operated by the University of Oregon and located near Bend, Oregon, (lat ~ 44 N, long ~ 121 W).
Open Zoom