1 ;;; GNU Guix --- Functional package management for GNU
2 ;;; Copyright © 2015, 2016, 2017 Ricardo Wurmus <rekado@elephly.net>
3 ;;; Copyright © 2016 Efraim Flashner <efraim@flashner.co.il>
4 ;;; Copyright © 2016, 2017 Marius Bakke <mbakke@fastmail.com>
5 ;;; Copyright © 2016 Hartmut Goebel <h.goebel@crazy-compilers.com>
6 ;;; Copyright © 2018 Tobias Geerinckx-Rice <me@tobias.gr>
8 ;;; This file is part of GNU Guix.
10 ;;; GNU Guix is free software; you can redistribute it and/or modify it
11 ;;; under the terms of the GNU General Public License as published by
12 ;;; the Free Software Foundation; either version 3 of the License, or (at
13 ;;; your option) any later version.
15 ;;; GNU Guix is distributed in the hope that it will be useful, but
16 ;;; WITHOUT ANY WARRANTY; without even the implied warranty of
17 ;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 ;;; GNU General Public License for more details.
20 ;;; You should have received a copy of the GNU General Public License
21 ;;; along with GNU Guix. If not, see <http://www.gnu.org/licenses/>.
23 (define-module (gnu packages machine-learning)
24 #:use-module ((guix licenses) #:prefix license:)
25 #:use-module (guix packages)
26 #:use-module (guix utils)
27 #:use-module (guix download)
28 #:use-module (guix svn-download)
29 #:use-module (guix build-system cmake)
30 #:use-module (guix build-system gnu)
31 #:use-module (guix build-system ocaml)
32 #:use-module (guix build-system python)
33 #:use-module (guix build-system r)
34 #:use-module (gnu packages)
35 #:use-module (gnu packages autotools)
36 #:use-module (gnu packages boost)
37 #:use-module (gnu packages check)
38 #:use-module (gnu packages compression)
39 #:use-module (gnu packages cran)
40 #:use-module (gnu packages dejagnu)
41 #:use-module (gnu packages gcc)
42 #:use-module (gnu packages image)
43 #:use-module (gnu packages maths)
44 #:use-module (gnu packages ocaml)
45 #:use-module (gnu packages perl)
46 #:use-module (gnu packages pkg-config)
47 #:use-module (gnu packages python)
48 #:use-module (gnu packages statistics)
49 #:use-module (gnu packages swig)
50 #:use-module (gnu packages xml)
51 #:use-module (gnu packages xorg))
60 (uri (string-append "https://www.csie.ntu.edu.tw/~cjlin/libsvm/"
61 name "-" version ".tar.gz"))
64 "0zd7s19y5vb7agczl6456bn45cj1y64739sslaskw1qk7dywd0bd"))))
65 (build-system gnu-build-system)
67 `(#:tests? #f ;no "check" target
68 #:phases (modify-phases %standard-phases
71 'install ; no ‘install’ target
72 (lambda* (#:key outputs #:allow-other-keys)
73 (let* ((out (assoc-ref outputs "out"))
74 (bin (string-append out "/bin/")))
76 (for-each (lambda (file)
77 (copy-file file (string-append bin file)))
82 (home-page "http://www.csie.ntu.edu.tw/~cjlin/libsvm/")
83 (synopsis "Library for Support Vector Machines")
85 "LIBSVM is a machine learning library for support vector
86 classification, (C-SVC, nu-SVC), regression (epsilon-SVR, nu-SVR) and
87 distribution estimation (one-class SVM). It supports multi-class
89 (license license:bsd-3)))
91 (define-public python-libsvm
92 (package (inherit libsvm)
93 (name "python-libsvm")
94 (build-system gnu-build-system)
96 `(#:tests? #f ;no "check" target
97 #:make-flags '("-C" "python")
99 (modify-phases %standard-phases
102 'install ; no ‘install’ target
103 (lambda* (#:key inputs outputs #:allow-other-keys)
104 (let ((site (string-append (assoc-ref outputs "out")
108 (assoc-ref inputs "python") 5) 3)
110 (substitute* "python/svm.py"
111 (("../libsvm.so.2") "libsvm.so.2"))
113 (for-each (lambda (file)
114 (copy-file file (string-append site (basename file))))
115 (find-files "python" "\\.py"))
116 (copy-file "libsvm.so.2"
117 (string-append site "libsvm.so.2")))
120 `(("python" ,python)))
121 (synopsis "Python bindings of libSVM")))
124 ;; The latest release candidate is several years and a couple of fixes have
125 ;; been published since. This is why we download the sources from the SVN
127 (let ((svn-revision 2341))
130 (version (string-append "0.9-rc3-0." (number->string svn-revision)))
134 (url "http://svn.code.sf.net/p/ghmm/code/trunk")
135 (revision svn-revision)))
136 (file-name (string-append name "-" version))
139 "0qbq1rqp94l530f043qzp8aw5lj7dng9wq0miffd7spd1ff638wq"))))
140 (build-system gnu-build-system)
142 `(#:imported-modules (,@%gnu-build-system-modules
143 (guix build python-build-system))
145 (modify-phases %standard-phases
146 (add-after 'unpack 'enter-dir
147 (lambda _ (chdir "ghmm") #t))
149 (add-after 'install 'check
150 (assoc-ref %standard-phases 'check))
151 (add-before 'check 'fix-PYTHONPATH
152 (lambda* (#:key inputs outputs #:allow-other-keys)
153 (let ((python-version ((@@ (guix build python-build-system)
155 (assoc-ref inputs "python"))))
157 (string-append (getenv "PYTHONPATH")
158 ":" (assoc-ref outputs "out")
159 "/lib/python" python-version
162 (add-after 'enter-dir 'fix-runpath
163 (lambda* (#:key outputs #:allow-other-keys)
164 (substitute* "ghmmwrapper/setup.py"
165 (("^(.*)extra_compile_args = \\[" line indent)
166 (string-append indent
167 "extra_link_args = [\"-Wl,-rpath="
168 (assoc-ref outputs "out") "/lib\"],\n"
171 (assoc-ref outputs "out")
174 (add-after 'enter-dir 'disable-broken-tests
176 (substitute* "tests/Makefile.am"
177 ;; GHMM_SILENT_TESTS is assumed to be a command.
178 (("TESTS_ENVIRONMENT.*") "")
179 ;; Do not build broken tests.
183 (("label_higher_order_test.*$")
184 "label_higher_order_test\n"))
186 ;; These Python unittests are broken as there is no gato.
187 ;; See https://sourceforge.net/p/ghmm/support-requests/3/
188 (substitute* "ghmmwrapper/ghmmunittests.py"
189 (("^(.*)def (testNewXML|testMultipleTransitionClasses|testNewXML)"
191 (string-append indent
192 "@unittest.skip(\"Disabled by Guix\")\n"
195 (add-after 'disable-broken-tests 'autogen
197 (zero? (system* "bash" "autogen.sh")))))))
199 `(("python" ,python-2) ; only Python 2 is supported
200 ("libxml2" ,libxml2)))
202 `(("pkg-config" ,pkg-config)
205 ("autoconf" ,autoconf)
206 ("automake" ,automake)
207 ("libtool" ,libtool)))
208 (home-page "http://ghmm.org")
209 (synopsis "Hidden Markov Model library")
211 "The General Hidden Markov Model library (GHMM) is a C library with
212 additional Python bindings implementing a wide range of types of @dfn{Hidden
213 Markov Models} (HMM) and algorithms: discrete, continuous emissions, basic
214 training, HMM clustering, HMM mixtures.")
215 (license license:lgpl2.0+))))
224 "http://micans.org/mcl/src/mcl-"
225 (string-replace-substring version "." "-")
229 "15xlax3z31lsn62vlg94hkm75nm40q4679amnfg13jm8m2bnhy5m"))))
230 (build-system gnu-build-system)
232 `(#:configure-flags (list "--enable-blast")))
235 (home-page "http://micans.org/mcl/")
236 (synopsis "Clustering algorithm for graphs")
238 "The MCL algorithm is short for the @dfn{Markov Cluster Algorithm}, a
239 fast and scalable unsupervised cluster algorithm for graphs (also known as
240 networks) based on simulation of (stochastic) flow in graphs.")
241 ;; In the LICENCE file and web page it says "The software is licensed
242 ;; under the GNU General Public License, version 3.", but in several of
243 ;; the source code files it suggests GPL3 or later.
244 ;; http://listserver.ebi.ac.uk/pipermail/mcl-users/2016/000376.html
245 (license license:gpl3)))
247 (define-public ocaml-mcl
250 (version "12-068oasis4")
255 "https://github.com/fhcrc/mcl/archive/"
257 (file-name (string-append name "-" version ".tar.gz"))
260 "1l5jbhwjpsj38x8b9698hfpkv75h8hn3kj0gihjhn8ym2cwwv110"))))
261 (build-system ocaml-build-system)
264 (modify-phases %standard-phases
265 (add-before 'configure 'patch-paths
267 (substitute* "configure"
268 (("SHELL = /bin/sh") (string-append "SHELL = "(which "sh"))))
269 (substitute* "setup.ml"
271 (string-append "LDFLAGS=-fPIC\"; \"SHELL=" (which "sh"))))
273 (home-page "https://github.com/fhcrc/mcl")
274 (synopsis "OCaml wrappers around MCL")
276 "This package provides OCaml bindings for the MCL graph clustering
278 (license license:gpl3)))
280 (define-public ocaml4.01-mcl
281 (package-with-ocaml4.01 ocaml-mcl))
283 (define-public randomjungle
285 (name "randomjungle")
291 "http://www.imbs-luebeck.de/imbs/sites/default/files/u59/"
292 "randomjungle-" version ".tar_.gz"))
295 "12c8rf30cla71swx2mf4ww9mfd8jbdw5lnxd7dxhyw1ygrvg6y4w"))))
296 (build-system gnu-build-system)
299 (list (string-append "--with-boost="
300 (assoc-ref %build-inputs "boost")))
302 (modify-phases %standard-phases
304 'configure 'set-CXXFLAGS
306 (setenv "CXXFLAGS" "-fpermissive ")
314 `(("gfortran" ,gfortran)
315 ("gfortran:lib" ,gfortran "lib")))
316 ;; Non-portable assembly instructions are used so building fails on
317 ;; platforms other than x86_64 or i686.
318 (supported-systems '("x86_64-linux" "i686-linux"))
319 (home-page "http://www.imbs-luebeck.de/imbs/de/node/227/")
320 (synopsis "Implementation of the Random Forests machine learning method")
322 "Random Jungle is an implementation of Random Forests. It is supposed to
323 analyse high dimensional data. In genetics, it can be used for analysing big
324 Genome Wide Association (GWA) data. Random Forests is a powerful machine
325 learning method. Most interesting features are variable selection, missing
326 value imputation, classifier creation, generalization error estimation and
327 sample proximities between pairs of cases.")
328 (license license:gpl3+)))
330 (define-public shogun
338 "ftp://shogun-toolbox.org/shogun/releases/"
339 (version-major+minor version)
340 "/sources/shogun-" version ".tar.bz2"))
343 "159nlijnb7mnrv9za80wnm1shwvy45hgrqzn51hxy7gw4z6d6fdb"))
344 (modules '((guix build utils)
348 ;; Remove non-free sources and files referencing them
349 (for-each delete-file
350 (find-files "src/shogun/classifier/svm/"
351 "SVMLight\\.(cpp|h)"))
352 (for-each delete-file
353 (find-files "examples/undocumented/libshogun/"
355 "(classifier_.*svmlight.*|"
356 "evaluation_cross_validation_locked_comparison).cpp")))
357 ;; Remove non-free functions.
358 (define (delete-ifdefs file)
359 (with-atomic-file-replacement file
361 (let loop ((line (read-line in 'concat))
363 (if (eof-object? line)
368 "#endif //USE_SVMLIGHT" line)))
370 "#ifdef USE_SVMLIGHT" line))))
371 (when (or (not skipping?)
372 (and skipping? (not skip-next?)))
374 (loop (read-line in 'concat) skip-next?)))))))
375 (for-each delete-ifdefs (find-files "src/shogun/kernel/"
376 "^Kernel\\.(cpp|h)"))))))
377 (build-system cmake-build-system)
379 '(#:tests? #f ;no check target
381 (modify-phases %standard-phases
382 (add-after 'unpack 'delete-broken-symlinks
384 (for-each delete-file '("applications/arts/data"
385 "applications/asp/data"
386 "applications/easysvm/data"
387 "applications/msplicer/data"
388 "applications/ocr/data"
389 "examples/documented/data"
390 "examples/documented/matlab_static"
391 "examples/documented/octave_static"
392 "examples/undocumented/data"
393 "examples/undocumented/matlab_static"
394 "examples/undocumented/octave_static"
395 "tests/integration/data"
396 "tests/integration/matlab_static"
397 "tests/integration/octave_static"
398 "tests/integration/python_modular/tests"))
400 (add-after 'unpack 'change-R-target-path
401 (lambda* (#:key outputs #:allow-other-keys)
402 (substitute* '("src/interfaces/r_modular/CMakeLists.txt"
403 "src/interfaces/r_static/CMakeLists.txt"
404 "examples/undocumented/r_modular/CMakeLists.txt")
405 (("\\$\\{R_COMPONENT_LIB_PATH\\}")
406 (string-append (assoc-ref outputs "out")
409 (add-after 'unpack 'fix-octave-modules
410 (lambda* (#:key outputs #:allow-other-keys)
411 (substitute* '("src/interfaces/octave_modular/CMakeLists.txt"
412 "src/interfaces/octave_static/CMakeLists.txt")
413 (("^include_directories\\(\\$\\{OCTAVE_INCLUDE_DIRS\\}")
414 "include_directories(${OCTAVE_INCLUDE_DIRS} ${OCTAVE_INCLUDE_DIRS}/octave"))
416 ;; change target directory
417 (substitute* "src/interfaces/octave_modular/CMakeLists.txt"
418 (("\\$\\{OCTAVE_OCT_LOCAL_API_FILE_DIR\\}")
419 (string-append (assoc-ref outputs "out")
420 "/share/octave/packages")))
422 (add-before 'build 'set-HOME
423 ;; $HOME needs to be set at some point during the build phase
424 (lambda _ (setenv "HOME" "/tmp") #t)))
426 (list "-DCMAKE_BUILD_WITH_INSTALL_RPATH=TRUE"
427 "-DUSE_SVMLIGHT=OFF" ;disable proprietary SVMLIGHT
428 ;;"-DJavaModular=ON" ;requires unpackaged jblas
429 ;;"-DRubyModular=ON" ;requires unpackaged ruby-narray
430 ;;"-DPerlModular=ON" ;"FindPerlLibs" does not exist
431 ;;"-DLuaModular=ON" ;fails because lua doesn't build pkgconfig file
438 "-DCmdLineStatic=ON")))
441 ("numpy" ,python-numpy)
442 ("r-minimal" ,r-minimal)
447 ("arpack" ,arpack-ng)
454 `(("pkg-config" ,pkg-config)))
455 ;; Non-portable SSE instructions are used so building fails on platforms
456 ;; other than x86_64.
457 (supported-systems '("x86_64-linux"))
458 (home-page "http://shogun-toolbox.org/")
459 (synopsis "Machine learning toolbox")
461 "The Shogun Machine learning toolbox provides a wide range of unified and
462 efficient Machine Learning (ML) methods. The toolbox seamlessly allows to
463 combine multiple data representations, algorithm classes, and general purpose
464 tools. This enables both rapid prototyping of data pipelines and extensibility
465 in terms of new algorithms.")
466 (license license:gpl3+)))
468 (define-public r-adaptivesparsity
470 (name "r-adaptivesparsity")
474 (uri (cran-uri "AdaptiveSparsity" version))
477 "1az7isvalf3kmdiycrfl6s9k9xqk22k1mc6rh8v0jmcz402qyq8z"))))
479 `((upstream-name . "AdaptiveSparsity")))
480 (build-system r-build-system)
483 (modify-phases %standard-phases
484 (add-after 'unpack 'link-against-armadillo
486 (substitute* "src/Makevars"
487 (("PKG_LIBS=" prefix)
488 (string-append prefix "-larmadillo"))))))))
491 ("r-rcpparmadillo" ,r-rcpparmadillo)))
493 `(("armadillo" ,armadillo)))
494 (home-page "https://cran.r-project.org/web/packages/AdaptiveSparsity")
495 (synopsis "Adaptive sparsity models")
497 "This package implements the Figueiredo machine learning algorithm for
498 adaptive sparsity and the Wong algorithm for adaptively sparse gaussian
500 (license license:lgpl3+)))
502 (define-public r-kernlab
509 (uri (cran-uri "kernlab" version))
512 "0qnaq9x3j2xc6jrmmd98wc6hkzch487s4p3a9lnc00xvahkhgpmr"))))
513 (build-system r-build-system)
514 (home-page "https://cran.r-project.org/web/packages/kernlab")
515 (synopsis "Kernel-based machine learning tools")
517 "This package provides kernel-based machine learning methods for
518 classification, regression, clustering, novelty detection, quantile regression
519 and dimensionality reduction. Among other methods @code{kernlab} includes
520 Support Vector Machines, Spectral Clustering, Kernel PCA, Gaussian Processes
522 (license license:gpl2)))
531 "http://dlib.net/files/dlib-" version ".tar.bz2"))
534 "1mljz02kwkrbggyncxv5fpnyjdybw2qihaacb3js8yfkw12vwpc2"))
535 (modules '((guix build utils)))
538 ;; Delete ~13MB of bundled dependencies.
539 (delete-file-recursively "dlib/external")
540 (delete-file-recursively "docs/dlib/external")
542 (build-system cmake-build-system)
545 (modify-phases %standard-phases
546 (add-after 'unpack 'disable-asserts
548 ;; config.h recommends explicitly enabling or disabling asserts
549 ;; when building as a shared library. By default neither is set.
550 (substitute* "dlib/config.h"
551 (("^//#define DLIB_DISABLE_ASSERTS") "#define DLIB_DISABLE_ASSERTS"))
553 (add-after 'disable-asserts 'disable-failing-tests
555 ;; One test times out on MIPS, so we need to disable it.
556 ;; Others are flaky on some platforms.
557 (let* ((system ,(or (%current-target-system)
559 (disabled-tests (cond
560 ((string-prefix? "mips64" system)
561 '("object_detector" ; timeout
563 ((string-prefix? "armhf" system)
564 '("learning_to_track"))
565 ((string-prefix? "i686" system)
570 (substitute* "dlib/test/makefile"
571 (((string-append "SRC \\+= " test "\\.cpp")) "")))
576 ;; No test target, so we build and run the unit tests here.
577 (let ((test-dir (string-append "../dlib-" ,version "/dlib/test")))
578 (with-directory-excursion test-dir
579 (and (zero? (system* "make" "-j" (number->string (parallel-job-count))))
580 (zero? (system* "./dtest" "--runall")))))))
581 (add-after 'install 'delete-static-library
582 (lambda* (#:key outputs #:allow-other-keys)
583 (delete-file (string-append (assoc-ref outputs "out")
587 `(("pkg-config" ,pkg-config)))
594 ("openblas" ,openblas)
597 "Toolkit for making machine learning and data analysis applications in C++")
599 "Dlib is a modern C++ toolkit containing machine learning algorithms and
600 tools. It is used in both industry and academia in a wide range of domains
601 including robotics, embedded devices, mobile phones, and large high performance
602 computing environments.")
603 (home-page "http://dlib.net")
604 (license license:boost1.0)))
606 (define-public python-scikit-learn
608 (name "python-scikit-learn")
614 "https://github.com/scikit-learn/scikit-learn/archive/"
616 (file-name (string-append name "-" version ".tar.gz"))
619 "18n8775kyfwbvcjjjzda9c5sqy4737c0hrmj6qj1ps2jmlqzair9"))
620 (patches (search-patches
621 "python-scikit-learn-fix-test-non-determinism.patch"))))
622 (build-system python-build-system)
625 (modify-phases %standard-phases
627 (add-after 'install 'check
628 ;; Running tests from the source directory requires
629 ;; an "inplace" build with paths relative to CWD.
630 ;; http://scikit-learn.org/stable/developers/advanced_installation.html#testing
631 ;; Use the installed version instead.
632 (lambda* (#:key inputs outputs #:allow-other-keys)
633 (add-installed-pythonpath inputs outputs)
634 ;; some tests require access to "$HOME"
635 (setenv "HOME" "/tmp")
636 ;; Step out of the source directory just to be sure.
638 (zero? (system* "nosetests" "-v" "sklearn")))))))
640 `(("openblas" ,openblas)))
642 `(("python-nose" ,python-nose)
643 ("python-cython" ,python-cython)))
645 `(("python-numpy" ,python-numpy)
646 ("python-scipy" ,python-scipy)))
647 (home-page "http://scikit-learn.org/")
648 (synopsis "Machine Learning in Python")
650 "Scikit-learn provides simple and efficient tools for data
651 mining and data analysis.")
652 (license license:bsd-3)))
654 (define-public python2-scikit-learn
655 (package-with-python2 python-scikit-learn))