version 1.3, 2014/03/24 07:54:51 |
version 1.12, 2016/02/05 01:58:34 |
Line 12 the Holonomic Gradient Descent Method (HGD) </h1> |
|
Line 12 the Holonomic Gradient Descent Method (HGD) </h1> |
|
|
|
<h2> Papers and Tutorials</h2> |
<h2> Papers and Tutorials</h2> |
<ol> |
<ol> |
<li> T.Koyama, A.Takemura, |
<li> Y.Goto, K.Matsumoto, |
|
Pfaffian equations and contiguity relations of the hypergeometric function of type (k+1,k+n+2) and their applications, |
|
<a href="http://arxiv.org/abs/1602.01637"> arxiv:1602.01637 |
|
|
|
<li>N.Marumo, T.Oaku, A.Takemura, |
|
Properties of powers of functions satisfying second-order linear differential equations with applications to statistics, |
|
<a href="http://arxiv.org/abs/1405.4451"> arxiv:1405.4451</a> |
|
|
|
<li> J.Hayakawa, A.Takemura, |
|
Estimation of exponential-polynomial distribution by holonomic gradient descent |
|
<a href="http://arxiv.org/abs/1403.7852"> arxiv:1403.7852</a> |
|
|
|
<li> C.Siriteanu, A.Takemura, S.Kuriki, |
|
MIMO Zero-Forcing Detection Performance Evaluation by Holonomic Gradient Method |
|
<a href="http://arxiv.org/abs/1403.3788"> arxiv:1403.3788</a> |
|
|
|
<li> T.Koyama, |
Holonomic Modules Associated with Multivariate Normal Probabilities of Polyhedra, |
Holonomic Modules Associated with Multivariate Normal Probabilities of Polyhedra, |
<a href="http://arxiv.org/abs/1311.6905"> arxiv:1311.6905 </a> |
<a href="http://arxiv.org/abs/1311.6905"> arxiv:1311.6905 </a> |
|
|
Line 36 Calculating the Normalising Constant of the Bingham Di |
|
Line 52 Calculating the Normalising Constant of the Bingham Di |
|
Statistics and Computing, 2013, |
Statistics and Computing, 2013, |
<a href="http://dx.doi.org/10.1007/s11222-013-9434-0">DOI</a> |
<a href="http://dx.doi.org/10.1007/s11222-013-9434-0">DOI</a> |
|
|
<li> T.Koyama, |
<li> T.Koyama, A.Takemura, |
Calculation of Orthant Probabilities by the Holonomic Gradient Method, |
Calculation of Orthant Probabilities by the Holonomic Gradient Method, |
<a href="http://arxiv.org/abs/1211.6822"> arxiv:1211.6822</a> |
<a href="http://arxiv.org/abs/1211.6822"> arxiv:1211.6822</a> |
|
|
<li>T. Koyama, H. Nakayama, K. Nishiyama, N. Takayama, |
<li>T. Koyama, H. Nakayama, K. Nishiyama, N. Takayama, |
Holonomic Rank of the Fisher-Bingham System of Differential Equations, |
Holonomic Rank of the Fisher-Bingham System of Differential Equations, |
<!-- <a href="http://arxiv.org/abs/1205.6144"> arxiv:1205.6144 </a>--> |
<!-- <a href="http://arxiv.org/abs/1205.6144"> arxiv:1205.6144 </a>--> |
to appear in Journal of Pure and Applied Algebra |
Journal of Pure and Applied Algebra (online), |
|
<a href="http://dx.doi.org/10.1016/j.jpaa.2014.03.004"> DOI </a> |
|
|
<li> |
<li> |
T. Koyama, H. Nakayama, K. Nishiyama, N. Takayama, |
T. Koyama, H. Nakayama, K. Nishiyama, N. Takayama, |
Line 66 Journal of Multivariate Analysis, 116 (2013), 440--455 |
|
Line 83 Journal of Multivariate Analysis, 116 (2013), 440--455 |
|
|
|
<li>T.Koyama, A Holonomic Ideal which Annihilates the Fisher-Bingham Integral, |
<li>T.Koyama, A Holonomic Ideal which Annihilates the Fisher-Bingham Integral, |
Funkcialaj Ekvacioj 56 (2013), 51--61. |
Funkcialaj Ekvacioj 56 (2013), 51--61. |
<!-- <a href="http://dx.doi.org/10.1619/fesi.56.51">DOI</a> --> |
<a href="http://dx.doi.org/10.1619/fesi.56.51">DOI</a> |
<a href="https://www.jstage.jst.go.jp/article/fesi/56/1/56_51/_article">jstage</a> |
<!-- <a href="https://www.jstage.jst.go.jp/article/fesi/56/1/56_51/_article">jstage</a> --> |
|
|
<li> |
<li> |
Hiromasa Nakayama, Kenta Nishiyama, Masayuki Noro, Katsuyoshi Ohara, |
Hiromasa Nakayama, Kenta Nishiyama, Masayuki Noro, Katsuyoshi Ohara, |
Line 80 Advances in Applied Mathematics 47 (2011), 639--658, |
|
Line 97 Advances in Applied Mathematics 47 (2011), 639--658, |
|
|
|
<h2> Three Steps of HGM </h2> |
<h2> Three Steps of HGM </h2> |
<ol> |
<ol> |
<li> Find a holonomic system satisfied by the normalizing constant. |
<li> Finding a holonomic system satisfied by the normalizing constant. |
We may use computational or theoretical methods to find it. |
We may use computational or theoretical methods to find it. |
Groebner basis and related methods are used. |
Groebner basis and related methods are used. |
<li> Find an initial value vector for the holonomic system. |
<li> Finding an initial value vector for the holonomic system. |
This is equivalent to evaluating the normalizing constant and its derivatives |
This is equivalent to evaluating the normalizing constant and its derivatives |
at a point. |
at a point. |
This step is usually performed by a series expansion. |
This step is usually performed by a series expansion. |
<li> Solve the holonomic system numerically. We use several methods |
<li> Solving the holonomic system numerically. We use several methods |
in numerical analysis such as the Runge-Kutta method of solving |
in numerical analysis such as the Runge-Kutta method of solving |
ordinary differential equations and efficient solvers of systems of linear |
ordinary differential equations and efficient solvers of systems of linear |
equations. |
equations. |
</ol> |
</ol> |
|
|
<h2> Software Packages for HGM</h2> |
<h2> Software Packages for HGM</h2> |
|
Most software packages are experimental and temporary documents are found in |
|
"asir-contrib manual" (auto-autogenerated part), or |
|
"Experimental Functions in Asir", or "miscellaneous and other documents" |
|
of the |
|
<a href="http://www.math.kobe-u.ac.jp/OpenXM/Current/doc/index-doc.html"> |
|
OpenXM documents</a> |
|
or in <a href="./"> this folder</a>. |
|
The nightly snapshot of the asir-contrib can be found in the asir page below, |
|
or look up our <a href="http://www.math.sci.kobe-u.ac.jp/cgi/cvsweb.cgi/"> |
|
cvsweb page</a>. |
<ol> |
<ol> |
<li> <a href="http://www.math.kobe-u.ac.jp/OpenXM/Math/hgm/"> hgm package for R </a> for the step 3. |
<li> Command line interfaces are in the folder OpenXM/src/hgm |
<li> yang (for Pfaffian systems) , nk_restriction (for D-module integrations), |
in the OpenXM source tree. See <a href="http://www.math.kobe-u.ac.jp/OpenXM"> |
tk_jack (for Jack polynomials) are for the steps 1 or 2 and in the |
OpenXM distribution page </a>. |
<a href="http://www.math.kobe-u.ac.jp/Asir"> asir-contrib </a> |
<li> <a href="http://www.math.kobe-u.ac.jp/OpenXM/Math/hgm/"> hgm package for R </a> (hgm_*tar.gz, hgm-manual.pdf) for the step 3. |
|
To install this package in R, type in |
|
<pre> |
|
R CMD install hgm_*.tar.gz |
|
</pre> |
|
<li> The following packages are |
|
for the computer algebra system |
|
<a href="http://www.math.kobe-u.ac.jp/Asir"> Risa/Asir</a>. |
|
They are in the asir-contrib collection. |
|
<ul> |
|
<li> yang.rr (for Pfaffian systems) , |
|
nk_restriction.rr (for D-module integrations), |
|
tk_jack.rr (for Jack polynomials), |
|
ko_fb_pfaffian.rr (Pfaffian system for the Fisher-Bingham system), |
|
are for the steps 1 or 2. |
|
<li> nk_fb_gen_c.rr is a package to generate a C program to perform |
|
maximal Likehood estimates for the Fisher-Bingham distribution by HGD (holonomic gradient descent). |
|
<li> ot_hgm_ahg.rr (HGM for A-distributions, very experimental). |
|
</ul> |
</ol> |
</ol> |
|
|
<h2> Programs to try examples of our papers </h2> |
<h2> Programs to try examples of our papers </h2> |
Line 106 tk_jack (for Jack polynomials) are for the steps 1 or |
|
Line 151 tk_jack (for Jack polynomials) are for the steps 1 or |
|
<li> <a href="http://www.math.kobe-u.ac.jp/OpenXM/Math/Fisher-Bingham-2"> d-dimensional Fisher-Bingham System </a> |
<li> <a href="http://www.math.kobe-u.ac.jp/OpenXM/Math/Fisher-Bingham-2"> d-dimensional Fisher-Bingham System </a> |
</ol> |
</ol> |
|
|
<pre> $OpenXM: OpenXM/src/hgm/doc/ref-hgm.html,v 1.2 2014/03/24 06:58:31 takayama Exp $ </pre> |
<pre> $OpenXM: OpenXM/src/hgm/doc/ref-hgm.html,v 1.11 2014/05/20 02:12:18 takayama Exp $ </pre> |
</body> |
</body> |
</html> |
</html> |