论文信息 - Software Packages for Holonomic Gradient Method

Software Packages for Holonomic Gradient Method

We present software packages for the holonomic gradient method (HGM). These packages compute normalizing constants and the probabilities of some regions. While many algorithms which compute in- tegrals over high-dimensional regions utilize the Monte-Carlo method, our HGM utilizes algorithms for solving ordinary differential equations such as the Runge-Kutta-Fehlberg method. As a result, our HGM can evaluate many integrals with a high degree of accuracy and moderate computational time. The source code of our packages is distributed on our web page (12).

[1] 李鎔範,et al. 数値計算のためのGNU Scientific Libraryの紹介(教育講座) , 2012 .

[2] Akimichi Takemura,et al. Calculation of orthant probabilities by the holonomic gradient method , 2012, ArXiv.

[3] Masayuki Noro,et al. Risa/Asir—a computer algebra system , 1992, ISSAC '92.

[4] D. Zeilberger. A holonomic systems approach to special functions identities , 1990 .

[5] Tomonari Sei,et al. Calculating the normalising constant of the Bingham distribution on the sphere using the holonomic gradient method , 2013, Stat. Comput..

[6] Anthony J. Hayter,et al. The evaluation of general non‐centred orthant probabilities , 2003 .

[7] Andrew T. A. Wood,et al. On the derivatives of the normalising constant of the Bingham distribution , 2007 .

[8] Radford M. Neal. Pattern Recognition and Machine Learning , 2007, Technometrics.

[9] Akimichi Takemura,et al. Holonomic Gradient Descent and its Application to Fisher-Bingham Integral , 2010, ArXiv.