Here we list student projects, past and present, and a short summary of the goals and achievements. The students listed have a greater or lesser Data Intensive component of their higher degree programs, but all are associated in one way or another.
We attempt to maintain a short description/title of their work and dates of commencing and finishing.
Current Students:
Xia Zhang
Kirstóf Rozgonyi
Title: Not finalized yet
Deep Investigation of Neutral Gas Origins, DINGO, is one of the ten ASKAP Survey Science Projects. The aim of DINGO is to study the role and evolution of HI from zero redshift up to z = 0.4 using 7500 hours of deep spectral line integration. Unfortunately with the current hardware support DINGO couldn't take full advantage of ASKAP and only deliver ~ 30'' resolution images. Furthermore as only the images of each observing session will be stored, the final data cube could be corrupted by low level interference, which is undetectable on a daily basis and span over the entire image. To overcome these problems an alternative approach of data storage required.
Storing the observed visibility would be ideal for dealing with low-level interference and improve the final image quality but this immensely increases output data size at the same time. One optimal solution is to store the visibility gridded on the uv plane. This method have the potential for data reduction due to the empty grid cells, thus the resolution of the survey could be increased to ~ 10'' including the long ( > 2 km) baselines. In addition image quality could be increased by handling the unforseen errors at the visibility domain.
Currently I am working on this problem simulating observations with different griddings to maximize the scientific improvement we can achieve on DINGO with storing the gridded uv visibility.
Geoff Dunaim
JT Malarecki
Kevin Vinsen
Ellie Gholami
Title: Automated Morphological Classification of Galaxies Using Deep Learning Approaches.
Nowadays, the morphological parameters of galaxies are of crucial requirements for studying galaxy formation and evolution.
Having access to many cosmological surveys such as GAMA and SDSS has opened a wide analysis of galaxy morphology.
The surveys are growing in terms of covered objects, making very hard and time-consuming the visual inspection of galaxies.
Automating the classification process has broadly come into the community. However, the efforts haven’t ended up with a high accuracy.
In this project, we will exploit neural network algorithms in order to classify a large number of galaxies. As our approach, we tend to use modified Convolutional Neural Network (CNN) and Galaxy Zoo as our initial training data set. Such a system would reduce the workload of
experts for classification. This algorithm would be applied to the large set of future surveys such as WAVES and LSST.
Manah Shah
David Gozzard
Ryan Bunney
Soheil Koushan
Completed Students:
Jason Wang
Jason was a Masters student working on GPU cluster based correlator, and then PhD student working on data I/O middleware. Now completed and working in the same area at the Oak Ridge National Laboratory in USA.
Derek Gerstmann
Derek was a PhD student, who did not complete the studies and returned to industry. His contributions however still continue to be used.
Chris Harris
Chris was one of the first PhD students to complete in ICRAR, and now works at the Pawsey centre.
Thanapol Chanapote
Thanapol was the first Thai student to visit under the MOU between ICRAR and NARIT. He was here for a year working on VLBI of OH masers, and now has completed his PhD. He has joined as a NARIT staff member.
Krystal Cook
Completed M.Sc. "Observing Radio Recombination Lines with the MWA".
Geoff Dunaim
Completed M.Phil. "Big Data Architecture in Radio Astronomy: The Effectiveness of the Hadoop/Hive/Spark ecosystem in Data Analysis of Large Astronomical Data Collections".