Randomized parallel computation
The goal is to demonstrate the power of randomization in designing efficient parallel algorithms for various computational problems. Specific problems of concern are: sorting, routing, and combinatorial optimization. A survey of existing randomized parallel algorithms for many important problems is given first. This is followed by a procedure for deriving randomized parallel algorithms for selection and sorting. Next, optimal algorithms are presented for sorting both general and integer keys. Also presented are algorithms that run in sub-logarithmic time. An optimal algorithm for permutation routing on a square mesh is given. A class of mesh-like networks that have optimal diameter are identified. Many heuristic algorithms have been proposed for solving presumably hard combinatorial problems. Simulated Annealing is one such heuristic that employs randomization; this procedures is modified for problems that have certain separation properties. This modified procedure (called nested Annealing) has a worst-case performance better than that of Simulated Annealing. Nested Annealing is shown to be easily parallelizable.
- Research Organization:
- Harvard Univ., Boston, MA (USA)
- OSTI ID:
- 6995544
- Country of Publication:
- United States
- Language:
- English
Similar Records
A Parallel Algorithm for the Vehicle Routing Problem
Mapping algorithms onto parallel architectures