Astronomy has well and truly entered a data-rich age, with observational surveys and computer simulations producing massive quantities of data. Traditional statistical and analysis methods are insufficient to truly exploit these giant datasets.  Machine learning, neural networks and deep learning methods are therefore poised to revolutionize the way we do astronomy.  ARCNet is an umbrella that encompasses three collaborations (GalNet, PlaNet and StarNet) within UVic’s Astronomy Research Centre (ARC) that are applying cutting edge data science techniques to astronomical datasets.  Students working in these groups, which have collaborations in computer science and the Canadian Astronomy Data Centre (CADC), are trained in both astronomy and machine learning techniques, which can be applied both within academia and in industry or the private sector.  ARCNet brings together these research groups under  the common goal of establishing UVic as a major centre for astronomy data science.

Info on upcoming ARCNet events can be found here.

For general inquiries contact:




The field of planetary science is undergoing a renaissance.  In the last decades, new regions of our solar system have been discovered, spacecraft have visited most of the planets of the solar system with more missions planned, the first planets outside the solar system have been discovered and directly imaged and new world telescopes, such as ALMA, are opening up new windows into understanding the early and late stages of the planet formation process.

This growth in planetary discoveries has been driven by complex data processing and data analytics. Surveys churn through petabytes of data in search of the smallest bodies in our solar system.  The mm-wavelength observed structures in disks of dust around stars reveal the complex processes of planetary formation via explorations of complex high-dimensionality parameter space.  The detection of planets in direct imaging is achieved through application of sophisticated numerical imaging processing techniques.

We are seeking students who are motivated to understand planets, planetary systems and the processes of planet formation via the application of digital technology tools to this complex data universe.  Come be a part of this revolution in understanding our place in the universe.

PlaNet lead:



We are entering an era of spectroscopic surveys for a wide range of science cases, including the ongoing Sloan Digital Sky Survey (SDSS) APOGEE and MaSTARS surveys, to the Gaia-ESO, DESI, LaMOST, WEAVE, 4MOST, PFS, and the future Maunakea Spectroscopic Explorer (MSE) surveys.  Stellar spectroscopic surveys provide a homogeneous database of stellar spectra that are ideal for advanced machine learning applications.

At UVic, we have developed a deep neural network to analyze both SDSS-III APOGEE data and synthetic stellar spectra.  Our neural network model (called StarNet) can produce stellar parameters with similar or better precision to the APOGEE pipeline, whether trained on APOGEE data or synthetic data.   The combination of increased computing resources and the availability of these large data sets are pivotal in the fast and efficient machine learning analysis of these data to maximize scientific impact. 

Our group seeks students interested in developing and improving StarNet.  We are also exploring its application to current and future optical spectral surveys of interest to astronomers around the world.   We believe in open source computing, and aim provide the resources that we develop for use by other astronomers and international facilities.  Our group is also exploring other advanced machine learning techniques, such as Generative Adversarial Networks.  New techniques like StarNet will be necessary to manage future large surveys, and they can have additional benefits such as the potential for real time applications to data taken on-sky.

StarNet lead:




The availability of massive public galaxy surveys, such as the Sloan Digital Sky Survey (SDSS), has revolutionized the power of archival research in extra-galacic astronomy.  Multi-wavelength complements, over large fractions of the sky, leverage incredible diagnostic power, at energies ranging from the UV (e.g. GALEX) to the mid-IR (e.g. WISE) all the way to the radio (e.g. ALFALFA).  This revolution is set to continue in the coming decades with even more massive datasets from projects such as Euclid and LSST.  In addition, projects with strong national leadership, such as the ongoing CFIS survey, and the future MSE facility, mean that Canadian astronomers are strategically positioned to engage in Big Data projects that tackle questions in extra-galactic astronomy.  A complementary approach can also be taken with public simulation datasets, such as Illustris and EAGLE, which also benefit from similar sophisticated analysis techniques as observational data.

The GalNet group at UVic already has a strong history in working with large galaxy survey data, across the electromagnetic spectrum and using simulations.  With expertize both in the Physics & Astronomy department, and collaborations in Computer Science and within NRC Herzberg, there is knowledge structure that spans astronomy, data science and cutting edge computational methods.  Recent projects include investigation of the processes that drive star formation quenching, predictions of galaxy gas masses and star formation rates using artificial neural networks and morphological classification of image data using deep learning.

GalNet lead: