The Investigation and Application of Tsallis Random Generator

• Main Authors: HsinShih Chen, 陳信實
• Other Authors: Jyhjeng Deng
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/03012760237855876804

碩士 === 大葉大學 === 工業工程學系碩士班 === 91 === Tsallis distribution was proposed by C. Tsallis in 1996 to solve the slow convergence problem of simulated annealing. It is shown that Tsallis’s generalized simulated annealing is much faster than the classical simulated annealing (“Boltzmann machine”) and fast simulated annealing (“Cauchy machine”). However, Tsallis distribution is very complicated and its random variable could not be generated by ordinary simulation techniques such as inversion and rejection methods. Tsallis adopts algorithm of R. N. Mantegna (1994) to produce a Tsallis random number generator. This method has many problems, however. First it could generate complex number when the parameter is near by 1.4. Second, when it is generated using Monte Carlo simulation, its histogram is not identical with the corresponding theoretical probability density (PDF). We plan to come out with a better Tsallis random number generator which can match the Tsallis’s PDF in most cases of its parameter’s ranges.