Machine learning of hierarchical clustering to segment n-dimensional images.
J. Nunez-Iglesias, R. Kennedy, T. Parag, J. Shi, and D. Chklovskii. Under Review. [arXiv]
Identifying maximal rigid components in bearing-based localization. R. Kennedy, K. Daniilidis, O. Naroditsky, and C.J. Taylor. IROS, 2012. [pdf] [slides] [video]
TopiaryExplorer: Visualizing large phylogenetic trees with environmental metadata. M. Pirrung, R. Kennedy, J.G. Caporaso, J. Stombaugh, D. Wendel, and R. Knight. Bioinformatics, 2011. [website] [online] [pdf] [PubMed]
Contour cut: identifying salient contours in images by solving a Hermitian eigenvalue problem. R. Kennedy, J. Gallier, and J. Shi. CVPR 2011. [pdf] [Supplementary Material] [Poster]
Natural and artificial RNAs occupy the same restricted region of sequence space. R. Kennedy, M. Lladser, Z. Wu, C. Zhang, M. Yarus, H. De Sterck, and R. Knight. RNA. 16:280-289, 2010.
Information, probability, and the abundance of the simplest RNA active sites. R. Kennedy, M. Lladser, M. Yarus, and R. Knight. Frontiers in Bioscience, 13:6060-6071, 2008.
Calculating RNA motif probabilities and recognizing patterns in sequence data. R. Kennedy. Senior Thesis, 2009.
Low-rank matrix completion. R. Kennedy. Written Preliminary Exam II, 2013.
Fall 2010 - TA for CIS520 Machine Learning
Spring 2011 - TA for CIS391 Introduction to Artificial Intelligence
Generalized distance transform
This is a simple MATLAB implmentation of the generalized distance transform algorithm from the paper Distance Transforms of Sampled Functions by P. Felzenszwalb and D. Huttenlocher. The function DT() gives the distance transform of a 2D image by calling DT1() for each dimension. By using DT1(), this could be easily extended to higher dimensions. It seems to have problems with inf values, so for points in the image with "no" parabola centered there, they should instead be given a large numeric value (such as 1e10). I also modified the algorithm so that the second argument returns the power diagram of the input. The power diagram is a diagram where each point is assigned to the point that is closest to it with respect to the distance transform. If all input points have the same value, this function reduces to giving the standard distance transform and the Voronoi diagram.