Time Series Center   Harvard University

We have developed an algorithm that allows fast and efficient detection of transits, including planetary transits, from light-curves. The method is based on building an ensemble of fiducial models and compressing the data using the MOPED algorithm. We will be searching for extra solar transit signatures in a huge collection of existing light curves (150 million light curves are in place at TSC right now).

more »

The detection of discrepant magnification events amongst the brightness records of multiple-image quasar have led to hypothesized models of observed variability in optical quasar light curves which include microlensing as a possible source of variation. Although an exciting possibility, it remains unclear whether potential microlensing events can be separated from intrinsic quasar variations. This project has explored a Haar wavelet based method designed to detect the presence or absence of microlensing effects in single-image quasar light curves based on temporal symmetry and asymmetry.

more »

Over the course of 3.5 year science mission, Pan-STARRS, will discover nearly every asteroid, Trojan, Centaur, long-period comet, short-period comet, and trans-neptunian object brighter than magnitude 23. This unprecedented census will be used to address a large number of questions regarding the physical and dynamical properties of the various small body populations of the solar system. Our project will use the output of this survey to determine the population of large, distant, and rare members of the outer solar system and and dynamically classify them.
Roughly 1-2% of TNOs are wide binaries with companions at separations greater than 1 arcsec and brightness differences less than 2 magnitudes. Based on the estimated 7000 TNOs that Pan STARRS will discover, we anticipate finding 70-140 wide binaries. The PS1 data, 60 epochs over three years, is naturally suited to determining the orbits of these objects. Our search will accurately determine the binary fraction for a variety of subclasses of TNOs.

more »

We have developed a de-trending algorithm for the removal of trends in time series. Trends in time series could be caused by various systematic and random noise sources such as cloud passages, changes of airmass, telescope vibration or CCD noise. Those trends undermine the intrinsic signals of stars and should be removed.

more »

We present 41 TAOS (the Taiwan-American Occulation Survey) Delta Sct Stars detected from TAOS 2-year data. Their amplitudes are a few tens of milli-magnitude and periods are about an hour, which strongly suggests that they are Low-Amplitude Delta Sct Stars (LADS). We also provide a revised catalog of all know Delta Sct Stars and their statistical properties.

more »

In the era of massive data sets, unexpected discoveries of astronomical phenomena through manual inspection of data is improbable if not impossible. Fortunately, automated 'anomaly' detection programs may be able to resurrect this mode of discovery and identify atypical phenomena indicative of novel astronomical objects.

We are interested in categorizing stars into known categories and finding stars whose brightness variation does not fit known categories, leading to new astronomical discoveries. We have developed PCAD, an unsupervised anomaly detection method that handles phase adjustments for large sets of periodic time-series data. The work needs to be extended to non-periodic data, and we want to develop clustering algorithms that are efficient in the face of noisy and voluminous data

more »

This project's role is to lay the foundations for providing research output from the TSC. We have developed a metadata (position, time, color, variability) database, across surveys, provides a useful compendium of temporal information on the brightness variability (or lack thereof) of most objects in the sky. We are currently developing multiple search interfaces to such a database, so that light curves satisfying arbitrary criterions may be searched for interesting objects. The searches will include morphological searches which compare light curves for similarity (DEMO coming SOON!). All searches will be subscribable, and web service outputs with and without light curves will be provided. A further challenge involves calibrating magnitudes across different color bands from different surveys. See the Search section of the website for current functionality.

more »

Data and search results are provided using a web interface as well as through web services. Additionally, summer 2007 intern Evan Morikawa has developed a filter and transformation workflow application FilterFlow which will enable structured and provenanced chains of filters to be developed in an example driven manner, either at the client or right at the server. In the era of multi-terabyte astronomy, shifting more of these pre-processing steps onto the server in structured and repeatable fashions will become increasingly important, as network bandwidth has not kept pace with storage capacity increases.

more »

Huge amounts of data are now available from astronomy surveys where telescopes take measurements of the sky over a long period of time. This information is processed by experts as well as computers to provide high-level classifications and measurements of importance for astronomy. Our project aims to provide automatic methods for classifying astronomical measurements of stars using luminosity time-series measurements and other measurable properties of the stars. Such classification systems can save many expert-human-years and allow for faster paced scientific investigation. Our research develops and applies machine learning techniques for this project including combining multiple forms of data (including time-series measurements), work on kernels, and providing interpretable results in an end to end prediction system. more »

This projects role is to make advanced research output products widely and quickly available. Feature space representations for light curves include spectral methods, wavelets, Singular Value Decomposition, segmentation methods and PAA, and discretization methods that transform the real-value time series into an alphabetic sequence (often reducing dimensionality as well) and include SAX and Persist. Databasing light curves by creating indexes using these representations allows for morphological, feature, or anomaly searches. We are developing streaming, multi-index database technology to make fast and efficient queries on very large sets of light curves.

more »

Wavelet decomposition has shown great promise in detecting signals from noise. One may decompose the original signal along with randomized sets of this signal into wavelet coefficients. The true signal can be differentiated from the randomized group. The detection is done due to the fact that wavelet coefficients of the true signal stand out 3 sigma away from the rest of the wavelet coefficients. The goal of this project is to utilize such methods in detecting true signals. more »

The discovery of events in time series can have important implications, such as identifying microlensing events in astronomical surveys, or changes in a patient's electrocardiogram. Current methods for identifying events require a fixed sliding window size, which is not ideal for all applications and could overlook important events. In this work, we develop probability models for finding the significance of an arbitrary-sized sliding window, and use these probabilities to find areas of significance. Because a brute force search of all sliding windows of all window sizes would be computationally intractable, we introduce a method for quickly approximating the results. We apply our method to our motivating domain of astronomy by analyzing over 100,000 time series from the MACHO survey, in which 56 different sections of the sky are considered, each with one or more known events. Our method was able to recover 100% of these events in the top 1% of the results, essentially pruning 99% of the data. Interestingly, our method was able to identify events that do not pass traditional event discovery procedures.

TAOS, MACHO.

more »