International Conference on Acoustics, Speech, and Signal Processing, 1984
Digital processing can be used to restore an image which has been blurred and/or corrupted by noi... more Digital processing can be used to restore an image which has been blurred and/or corrupted by noise. This is a very slow process when implemented on a general purpose computer, especially for non-homogeneous problems, where FFT methods cannot be used. This paper describes an algorithm and a processor architecture for restoring an image in parallel, one pixel row at a
ABSTRACT LU factorization of matrices with irregular sparsity patterns is a challenging applicati... more ABSTRACT LU factorization of matrices with irregular sparsity patterns is a challenging application for parallel processors. We have found that the data-parallel programming model can provide very high performance for these computations. This paper presents a new, efficient family of data mappings, and performance measurements for the MasPar MP-2 and the Thinking Machines CM-5 on several standard benchmark problems, showing that our approach is very competitive with message-passing implementations as well as vector supercomputer codes.
Abstract: It is becoming increasingly popular to represent myriad and diverse data sets as graphs... more Abstract: It is becoming increasingly popular to represent myriad and diverse data sets as graphs. When the labels of vertices of these graphs are unavailable, graph matching (GM)---the process of determining which permutation assigns vertices of one graph to those of another---is a computationally daunting problem. This work presents an inexact strategy for GM. Specifically, we frame GM as a quadratic assignment problem, and then relax the feasible region to its convex hull. We prove that our relaxed optimization function has the ...
Research is being conducted on high power microwave devices (eg, gyrotrons) at the University of ... more Research is being conducted on high power microwave devices (eg, gyrotrons) at the University of Michigan. Of utmost concern is the phenomenon of pulse shortening, that is, the duration of the microwave pulse is shorter than the duration of the cathode voltage. For years researchers have applied the Fourier transform to the heterodyned microwave signals. The problem with this technique is that a signal with multiple frequency components has the same spectrum as that of a signal with frequency components emitted at different times. Time-frequency ...
Quadratic assignment problems arise in a wide variety of domains, spanning operations research, g... more Quadratic assignment problems arise in a wide variety of domains, spanning operations research, graph theory, computer vision, and neuroscience, to name a few. The graph matching problem is a special case of the quadratic assignment problem, and graph matching is increasingly important as graph-valued data is becoming more prominent. With the aim of efficiently and accurately matching the large graphs common in big data, we present our graph matching algorithm, the Fast Approximate Quadratic assignment algorithm. We empirically demonstrate that our algorithm is faster and achieves a lower objective value on over 80% of the QAPLIB benchmark library, compared with the previous state-of-the-art. Applying our algorithm to our motivating example, matching C. elegans connectomes (brain-graphs), we find that it efficiently achieves performance.
International Conference on Acoustics, Speech, and Signal Processing, 1984
Digital processing can be used to restore an image which has been blurred and/or corrupted by noi... more Digital processing can be used to restore an image which has been blurred and/or corrupted by noise. This is a very slow process when implemented on a general purpose computer, especially for non-homogeneous problems, where FFT methods cannot be used. This paper describes an algorithm and a processor architecture for restoring an image in parallel, one pixel row at a
ABSTRACT LU factorization of matrices with irregular sparsity patterns is a challenging applicati... more ABSTRACT LU factorization of matrices with irregular sparsity patterns is a challenging application for parallel processors. We have found that the data-parallel programming model can provide very high performance for these computations. This paper presents a new, efficient family of data mappings, and performance measurements for the MasPar MP-2 and the Thinking Machines CM-5 on several standard benchmark problems, showing that our approach is very competitive with message-passing implementations as well as vector supercomputer codes.
Abstract: It is becoming increasingly popular to represent myriad and diverse data sets as graphs... more Abstract: It is becoming increasingly popular to represent myriad and diverse data sets as graphs. When the labels of vertices of these graphs are unavailable, graph matching (GM)---the process of determining which permutation assigns vertices of one graph to those of another---is a computationally daunting problem. This work presents an inexact strategy for GM. Specifically, we frame GM as a quadratic assignment problem, and then relax the feasible region to its convex hull. We prove that our relaxed optimization function has the ...
Research is being conducted on high power microwave devices (eg, gyrotrons) at the University of ... more Research is being conducted on high power microwave devices (eg, gyrotrons) at the University of Michigan. Of utmost concern is the phenomenon of pulse shortening, that is, the duration of the microwave pulse is shorter than the duration of the cathode voltage. For years researchers have applied the Fourier transform to the heterodyned microwave signals. The problem with this technique is that a signal with multiple frequency components has the same spectrum as that of a signal with frequency components emitted at different times. Time-frequency ...
Quadratic assignment problems arise in a wide variety of domains, spanning operations research, g... more Quadratic assignment problems arise in a wide variety of domains, spanning operations research, graph theory, computer vision, and neuroscience, to name a few. The graph matching problem is a special case of the quadratic assignment problem, and graph matching is increasingly important as graph-valued data is becoming more prominent. With the aim of efficiently and accurately matching the large graphs common in big data, we present our graph matching algorithm, the Fast Approximate Quadratic assignment algorithm. We empirically demonstrate that our algorithm is faster and achieves a lower objective value on over 80% of the QAPLIB benchmark library, compared with the previous state-of-the-art. Applying our algorithm to our motivating example, matching C. elegans connectomes (brain-graphs), we find that it efficiently achieves performance.
Uploads
Papers by Steven Kratzer