hide You are viewing an archived web page, collected at the request of Harvard University Archives using Archive-It. This page was captured on 00:26:43 Apr 22, 2014, and is part of the A-Sites: Archived Harvard Websites collection. The information on this web page may be out of date. See All versions of this archived page. Loading media information


Hands-on learning through internships and Research Experiences for Undergraduates

Intern presentations

Videos will be posted when available.


  • Samuel Chen: Using Wavelet Decomposition for Signal Detection
  • Kevin Dale: Leveraging Commodity Graphics Hardware for Radio Astronomy
  • Robert Harris: Employing Medical Imaging Tools to Understand High Mass Star Formation
  • David Kosslyn: The Time Series Center
  • Evan Morikawa: A Novel GUI Based Interactive Work Flow Application for Exploratory and Batch Processing of Light Curves


  • Tom Buckley: Collaboration Within the IIC
  • Mark Goetz: Graphical Viewer for Biological Ontologies in StemBook
  • Mike Horn: Evolve: Exploring Biodiversity
  • Hao Jiang: CThru: Exploration in a Video-Centered Information Space for Educational Purposes
  • Dae-Won Kim: Removal of Trends in Astronomy Time Series
  • Christian Lederberger: Volume MLS Raycasting
  • Amelio Vázquez-Reina: Active Ribbons for the Connectome
  • Gabriel Wachman: Automatic Classification of Variable Stars

Student posters

Presented at the IIC Open House, 2008

  • Kevin Dale: Lab Trials in GPUs for the Murchison Widefield Array
The use of Graphics Processing Units (GPUs) for non-graphics applications represents a growing trend, one driven by the tremendous floating point capability and relatively modest price of commodity graphics processors. This work explores the suitability of GPUs for real-time data processing for the Murchison Widefield Array (MWA) radio telescopes. Our single-GPU implementation of the major stages of array calibration and image formation provides an average speedup over a single CPU of about 10x, with more than a 60x speedup for the most improved stage. Our new algorithm for global removal of trends in time series is based on an algorithm originally developed for removing trends, such as weather changes, that appear in time series data of star brightness. The algorithm we have developed can be applied to any time series data that show trends. The algorithm is based on a Pearson correlation matrix of all data sets. We determine trends by summing the subset of datasets that are strongly correlated among themselves. A clustering algorithm is used to extract those highly correlated subsets. Experimental results with simulated data are presented. We also applied our algorithm to stock-market data.
  • Miriah Meyer: Visualizing Gene Duplications in the Rhizopus Genome
By studying genes that appear to be duplicated across the genome of the fungus Rhizopus oryzae, scientist Li-Jun Ma at the Broad Institute is working to unlock the secrets of the organism's ability to quickly divide, a characteristic that makes the fungus extremely dangerous to humans when infiltrated into the blood stream. Visualizing known gene duplication sites in conjunction with transposons and tRNA also present in the genome allows Ma to extrapolate that larger blocks of genomic region are in fact duplicated, revealing insight into the ancient evolutionary history of this organism. To aid in the understanding of the relationships between these genetic elements, we present an intuitive visualization of Dr. Ma's data that allows the user to control the amount of data presented such that small pieces of the genome can be studied more closely. 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 sliding window of a constrained size, which is not ideal for all applications and could cause the scanner to 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 applied our method to our motivating domain of astronomy by analyzing over 500,000 time series from the MACHO survey.
  • Umaa Rebbapragada: Outlier Detection in a Set of Time Series
  • Gabriel Wachman: Classifying Variable Stars Using Machine Learning Methods
Our goal is to automate the classification of variable stars. To accomplish this, we have developed a system that uses machine learning methods to identify stars as periodic and classify them as Eclipsing Binary, RRL or Cepheid. We hope to publish results on the MACHO catalog in the near future. We plan to use the ASAS and TYCHO catalogs as training sets in order to be able to classify other unlabeled catalogs.


Document Actions
contact information


Rosalind Reid