1 2# STK: a Small (Matlab/Octave) Toolbox for Kriging 3 4This README file is part of 5 6*STK: a Small (Matlab/Octave) Toolbox for Kriging* 7<http://sourceforge.net/projects/kriging> 8 9STK is free software: you can redistribute it and/or modify it under 10the terms of the GNU General Public License as published by the Free 11Software Foundation, either version 3 of the License, or (at your 12option) any later version. 13 14STK is distributed in the hope that it will be useful, but WITHOUT 15ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 16or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public 17License for more details. 18 19You should have received a copy of the GNU General Public License 20along with STK. If not, see <http://www.gnu.org/licenses/>. 21 22 23## General information 24 25 Version: 2.6.1 (2019-10-12) 26 27 Authors: See AUTHORS.md file 28 29 Maintainers: Julien Bect <julien.bect@centralesupelec.fr> 30 and Emmanuel Vazquez <emmanuel.vazquez@centralesupelec.fr> 31 32 Description: The STK is a (not so) Small Toolbox for Kriging. Its 33 primary focus is on the interpolation/regression 34 technique known as kriging, which is very closely related 35 to Splines and Radial Basis Functions, and can be 36 interpreted as a non-parametric Bayesian method using a 37 Gaussian Process (GP) prior. The STK also provides tools 38 for the sequential and non-sequential design of 39 experiments. Even though it is, currently, mostly geared 40 towards the Design and Analysis of Computer Experiments 41 (DACE), the STK can be useful for other applications 42 areas (such as Geostatistics, Machine Learning, 43 Non-parametric Regression, etc.). 44 45 Copyright: Large portions are Copyright (C) 2011-2014 SUPELEC 46 and Copyright (C) 2015-2019 CentraleSupelec. 47 See individual copyright notices for more details. 48 49 License: GNU General Public License, version 3 (GPLv3). 50 See COPYING for the full license. 51 52 URL: <http://sourceforge.net/projects/kriging> 53 54 55## One toolbox, two flavours 56 57The STK toolbox comes in two flavours: 58 59 * an "all purpose" release, which is suitable for use both with 60 [GNU Octave](http://www.gnu.org/software/octave/) 61 and with [Matlab](www.mathworks.com/products/matlab/). 62 * an Octave package, for people who want to install and use STK as a 63 regular [Octave package](http://www.gnu.org/software/octave/doc/interpreter/Packages.html#Packages). 64 65Hint: if you're not sure about the version that you have... 66 67 * the "all purpose" release has this file (`README.md`) and the `stk_init` 68 function (`stk_init.m`) in the top-level directory, 69 * the Octave package has a `DESCRIPTION` file in the top-level directory 70 and this file in the `doc/` subdirectory. 71 72 73## Quick Start 74 75### Quick start with the "all purpose" release (Matlab/Octave) 76 77Download and unpack an archive of the "all purpose" release from the 78[STK project](https://sourceforge.net/projects/kriging/) 79[file release system](https://sourceforge.net/projects/kriging/files/) 80on SourceForge. 81 82Run `stk_init.m` in either Octave or Matlab. 83 84After that, you should be able to run the examples located in the `examples` 85directory. All of them are scripts, the file name of which starts with 86the `stk_example_` prefix. 87 88For instance, type `stk_example_kb03` to run the third example in the "Kriging 89basics" series. 90 91### Quick start with the Octave package release (Octave only) 92 93Assuming that you have a working Internet connection, typing `pkg install -forge stk` 94(from within Octave) will automatically download the latest STK package tarball from the 95[Octave Forge](http://octave.sourceforge.net/stk/index.html) 96[file release system](https://sourceforge.net/projects/octave/files/) 97on SourceForge and install it for you. 98 99Alternatively, if you want to install an older (or beta) release, you can download 100the tarball from either the STK project FRS or the Octave Forge FRS, and install it 101with `pkg install FILENAME.tar.gz`. 102 103After that, you can load STK using `pkg load stk`. 104 105To check that STK is properly loaded, try for instance `stk_example_kb03` to run 106the third example in the "Kriging basics" series. 107 108 109## Requirements and recommendations 110 111### Common requirement 112 113 Your installation must be able to compile C mex files. 114 115### Requirements and recommendations for use with GNU Octave 116 117 The STK is tested to work with GNU Octave 3.8.2 or newer, but 118 should probably also work with Octave 3.8.0 and 3.8.1. 119 120 Older versions of Octave (<= 3.6.2) are no longer supported, and 121 are known to contain bugs that prevent some STK functions from 122 working properly. 123 124 125### Requirements and recommendations for use with Matlab 126 127 The STK works with Matlab R2007a or newer. 128 129 The Optimization Toolbox is recommended. 130 131 The Parallel Computing Toolbox is optional. 132 133 134## Content 135 136 By publishing this toolbox, the idea is to provide a convenient and 137 flexible research tool for working with kriging-based methods. The 138 code of the toolbox is meant to be easily understandable, modular, 139 and reusable. By way of illustration, it is very easy to use this 140 toolbox for implementing the EGO algorithm [1]. 141 Besides, this toolbox can serve as a basis for the implementation 142 of advanced algorithms such as Stepwise Uncertainty Reduction (SUR) 143 algorithms [2]. 144 145 The toolbox consists of three parts: 146 147 1. The first part is the implementation of a number of covariance 148 functions, and tools to compute covariance vectors and matrices. 149 The structure of the STK makes it possible to use any kind of 150 covariances: stationary or non-stationary covariances, aniso- 151 tropic covariances, generalized covariances, etc. 152 153 2. The second part is the implementation of a REMAP procedure to 154 estimate the parameters of the covariance. This makes it possible 155 to deal with generalized covariances and to take into account 156 prior knowledge about the parameters of the covariance. 157 158 3. The third part consists of prediction procedures. In its current 159 form, the STK has been optimized to deal with moderately large 160 data sets. 161 162### References 163 164[1] D. R. Jones, M. Schonlau, and William J. Welch. *Efficient global 165optimization of expensive black-box functions*. Journal of Global 166Optimization, 13(4):455-492, 1998. 167 168[2] J. Bect, D. Ginsbourger, L. Li, V. Picheny, and E. Vazquez. 169*Sequential design of computer experiments for the estimation of a 170probability of failure*. Statistics and Computing, pages 1-21, 2011. 171DOI: 10.1007/s11222-011-9241-4. 172 173 174## Ways to get help, report bugs, ask for new features... 175 176 Use the "help" mailing-list: 177 178 <kriging-help@lists.sourceforge.net> 179 <https://sourceforge.net/p/kriging/mailman> 180 181 to ask for help on STK, and the ticket manager: 182 183 <https://sourceforge.net/p/kriging/tickets> 184 185 to report bugs or ask for new features. 186