Joseph RichardsPostdoctoral Researcher
University of California at Berkeley
Center for Time Domain Informatics
481 Evans Hall
Berkeley, CA 94720
***NEW*** Active Learning Paper Published in ApJ!
My area of interest is Astrostatistics. Modern-day astronomical
surveys produce massive data streams, requiring sophisticated
statistical analysis. Currently, my main area of focus is
classification of sparse and noisy astronomical time series.
Upcoming photometric surveys such as LSST will observe ~105
new transient and variable sources per night. Fast and
accurate classification tools are needed to determine the
nature of each new source and to find the most interesting
objects for follow-up observations and further astrophysical
analysis. I am involved with several projects here in the
Center for Time Domain Informatics to develop statistical
methodology for this classification problem and implement
these methods for the Palomar Transient
For my Ph.D. work, I introduced the diffusion map, a method of non-linear dimensionality reduction, to astrophysics, applying it to such diverse problems as star formation history estimation, photometric-redshift prediction, and supernova light curve typing. These methods are useful in a variety of problems, and I am currently researching the use of diffusion map to classify astronomical transients and variable stars. For a list of my research interests and projects, please visit my Research page or my CV.
Some Recent Publications
Richards, Joseph W.; Lee, Ann B.; Schafer, Chad M.; Freeman, Peter E.. (2011) Prototype Selection for Parameter Estimation in Complex Models. Annals of Applied Statistics 2012, Vol. 6, No. 1, 383-408. arXiv:1105.6344