Carla MartinEmail: carlam Abstract on my research: Tensor Decompositions Suppose you have two digital images of yourself. In one image you are smiling at the camera, but in the other image you are yawning and your face is turned slightly away from the camera. How can a computer identify that your two images represent the same person? This is just one application of tensor decompositions, which is a specialty area of numerical linear algebra. Typically, algorithms involving matrix computations take advantage of matrix decompositions in order to perform computations more efficiently. Imagine that instead of representing data in a matrix (a square or rectangular viewpoint), it is represented as a rectangular cube of data. My research involves computing certain decompositions of these cubes, or tensors, of data. These decompositions are used in computer image recognition, statistics, medical imaging, chemistry, and many other applications. |
cam.cornell.edu