PI: Erik Katsavounidis, Department of Physics, MIT
Physics and Astronomy projects have been prodigious in generating data well before the official Big Data term was coined. With data archives to process at the petabyte-scale (per year) and petaflop-scale computing needed, gravitational-wave observatories, but also astronomical telescopes scanning the sky for the next astrophysical explosion or a near-Earth asteroid have been among such projects. Following the first direct observation of gravitational waves by LIGO in 2015, the synergy between gravitational- and electromagnetic-wave astronomy will be the next and necessary step in re-defining the way we do astrophysics and learn about the Cosmos. This proposal brings together expertise in Big Data science, as well gravitational- and electromagnetic-wave astronomy in order to work toward this connection, address the computational challenges and enable an international partnership that will maximize the science and its applications in a multi-disciplinary way.
- Project Title: Big Data Challenges and International Collaborative Programs inGravitational-Wave Physics and Astronomy
- Principal Investigator: Erik Katsavounidis, Department of Physics, MIT
- Grant Period: February 2017 – January 2018
In early summer of 2017 we had the first visit by Skoltech's Evgeny Burnaev, associate professor at the Center for Computational Data-Intensive Science and Engineering (June 29- July 5). This was within the context of a computing workshop on the future of high performance computing Erik Katsavounidis was co-organizing (www.mghpcc.org/hpc-futures/). Evgeny participated in the workshop and held meetings with colleagues at MIT both in EECS and physics, including Erik Katsavounidis and members of his group (Ryan Lynch, graduate student, Satya Mohapatra, staff scientist and Denis Martynov, postdoc).
Meeting discussions in this first meeting focused on two main directions: (1) anomaly detection in the presence of non-Gaussian, non-stationary noise. Erik presented the basic signal processing challenge we face in the analysis of data collected by the kilometer-scale ground interferometers for gravitational-wave detection. Evgeny presented the analogous problem his group is tackling and we tried to draw similarities and cross-application of such transient-detection algorithms. And (2) we developed an idea for "University Cloud" by federating idling computing resources within a university. Discussed platforms for implementing such solution, including ways to incorporate encryption.
Evgeny's summer visit was followed by a visit to Moscow of Ryan Lynch, a graduate student working with Erik on transient detection. Ryan was in Moscow the first week of September 2017 and spent time at Skoltech within Maxim Fedorov's and Evgeny Burnaev's group as well as Dmitry Bisikalo's group at INASAN. Ryan presented talks at Skoltech and INASAN covering the physics and signal processing aspects of our data analysis work. He met with graduate students working on similar problems at Skoltech. The interaction led to identifying a common dataset in order to benchmark methods, including developing standards for comparison. As part of that we have prepared a small sample of instrument (LIGO) data in the NumPy file format (rather than the proprietary data structures we otherwise use) in order to allow easy access of such data to the Skoltech group. This sample of data is planned to be used for benchmarking and improving our anomaly detection algorithms.
In late November 2017, we organized a 2-day workshop at MIT on the design of an optical telescope multi-purpose network. The meeting brought to MIT the director of INASAN and co-I in this proposal Dmitry Bisikalo as well as INASAN scientists Andrey Shukharov and Igor Savanov. MIT physics faculty Rob Simcoe and Salvo Vitale, as well as Harvard physics faculty Chris Stubbs participated in this workshop. Our INASAN colleagues presented new designs for optical telescopes that can support the science with gravitational-wave astrophysics, but also provide other applications with broader interest like space debris and asteroid/comets prompt identification. They have identified and we discussed design options and trade offs between costs and science capacity. On our end, we presented the set of requirements from the gravitational-wave perspective. The workshop ended with good material in hand in order to pursue formal proposal writing with funding agencies in the US and Russia.
In the upcoming six months, we would like to complete the study on anomaly detection with the data set we made available as well as collaborate further on proposal opportunities for the construction of an telescope network. We anticipate additional personnel visits between our sites and as our funding permits.