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>
7 ;;; This file is part of GNU Guix.
9 ;;; GNU Guix is free software; you can redistribute it and/or modify it
10 ;;; under the terms of the GNU General Public License as published by
11 ;;; the Free Software Foundation; either version 3 of the License, or (at
12 ;;; your option) any later version.
14 ;;; GNU Guix is distributed in the hope that it will be useful, but
15 ;;; WITHOUT ANY WARRANTY; without even the implied warranty of
16 ;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 ;;; GNU General Public License for more details.
19 ;;; You should have received a copy of the GNU General Public License
20 ;;; along with GNU Guix. If not, see <http://www.gnu.org/licenses/>.
22 (define-module (gnu packages machine-learning)
23 #:use-module ((guix licenses) #:prefix license:)
24 #:use-module (guix packages)
25 #:use-module (guix utils)
26 #:use-module (guix download)
27 #:use-module (guix svn-download)
28 #:use-module (guix build-system cmake)
29 #:use-module (guix build-system gnu)
30 #:use-module (guix build-system ocaml)
31 #:use-module (guix build-system python)
32 #:use-module (guix build-system r)
33 #:use-module (gnu packages)
34 #:use-module (gnu packages autotools)
35 #:use-module (gnu packages boost)
36 #:use-module (gnu packages check)
37 #:use-module (gnu packages compression)
38 #:use-module (gnu packages cran)
39 #:use-module (gnu packages dejagnu)
40 #:use-module (gnu packages gcc)
41 #:use-module (gnu packages image)
42 #:use-module (gnu packages maths)
43 #:use-module (gnu packages ocaml)
44 #:use-module (gnu packages perl)
45 #:use-module (gnu packages pkg-config)
46 #:use-module (gnu packages python)
47 #:use-module (gnu packages statistics)
48 #:use-module (gnu packages swig)
49 #:use-module (gnu packages xml)
50 #:use-module (gnu packages xorg))
60 "https://github.com/cjlin1/libsvm/archive/v"
61 (string-delete #\. version) ".tar.gz"))
62 (file-name (string-append name "-" version ".tar.gz"))
65 "1jpjlql3frjza7zxzrqqr2firh44fjb8fqsdmvz6bjz7sb47zgp4"))))
66 (build-system gnu-build-system)
68 `(#:tests? #f ;no "check" target
69 #:phases (modify-phases %standard-phases
73 (lambda* (#:key outputs #:allow-other-keys)
74 (let* ((out (assoc-ref outputs "out"))
75 (bin (string-append out "/bin/")))
77 (for-each (lambda (file)
78 (copy-file file (string-append bin file)))
83 (home-page "http://www.csie.ntu.edu.tw/~cjlin/libsvm/")
84 (synopsis "Library for Support Vector Machines")
86 "LIBSVM is a machine learning library for support vector
87 classification, (C-SVC, nu-SVC), regression (epsilon-SVR, nu-SVR) and
88 distribution estimation (one-class SVM). It supports multi-class
90 (license license:bsd-3)))
92 (define-public python-libsvm
93 (package (inherit libsvm)
94 (name "python-libsvm")
95 (build-system gnu-build-system)
97 `(#:tests? #f ;no "check" target
98 #:make-flags '("-C" "python")
100 (modify-phases %standard-phases
104 (lambda* (#:key inputs outputs #:allow-other-keys)
105 (let ((site (string-append (assoc-ref outputs "out")
109 (assoc-ref inputs "python") 5) 3)
111 (substitute* "python/svm.py"
112 (("../libsvm.so.2") "libsvm.so.2"))
114 (for-each (lambda (file)
115 (copy-file file (string-append site (basename file))))
116 (find-files "python" "\\.py"))
117 (copy-file "libsvm.so.2"
118 (string-append site "libsvm.so.2")))
121 `(("python" ,python)))
122 (synopsis "Python bindings of libSVM")))
125 ;; The latest release candidate is several years and a couple of fixes have
126 ;; been published since. This is why we download the sources from the SVN
128 (let ((svn-revision 2341))
131 (version (string-append "0.9-rc3-0." (number->string svn-revision)))
135 (url "http://svn.code.sf.net/p/ghmm/code/trunk")
136 (revision svn-revision)))
137 (file-name (string-append name "-" version))
140 "0qbq1rqp94l530f043qzp8aw5lj7dng9wq0miffd7spd1ff638wq"))))
141 (build-system gnu-build-system)
143 `(#:imported-modules (,@%gnu-build-system-modules
144 (guix build python-build-system))
146 (modify-phases %standard-phases
147 (add-after 'unpack 'enter-dir
148 (lambda _ (chdir "ghmm") #t))
150 (add-after 'install 'check
151 (assoc-ref %standard-phases 'check))
152 (add-before 'check 'fix-PYTHONPATH
153 (lambda* (#:key inputs outputs #:allow-other-keys)
154 (let ((python-version ((@@ (guix build python-build-system)
156 (assoc-ref inputs "python"))))
158 (string-append (getenv "PYTHONPATH")
159 ":" (assoc-ref outputs "out")
160 "/lib/python" python-version
163 (add-after 'enter-dir 'fix-runpath
164 (lambda* (#:key outputs #:allow-other-keys)
165 (substitute* "ghmmwrapper/setup.py"
166 (("^(.*)extra_compile_args = \\[" line indent)
167 (string-append indent
168 "extra_link_args = [\"-Wl,-rpath="
169 (assoc-ref outputs "out") "/lib\"],\n"
172 (assoc-ref outputs "out")
175 (add-after 'enter-dir 'disable-broken-tests
177 (substitute* "tests/Makefile.am"
178 ;; GHMM_SILENT_TESTS is assumed to be a command.
179 (("TESTS_ENVIRONMENT.*") "")
180 ;; Do not build broken tests.
184 (("label_higher_order_test.*$")
185 "label_higher_order_test\n"))
187 ;; These Python unittests are broken as there is no gato.
188 ;; See https://sourceforge.net/p/ghmm/support-requests/3/
189 (substitute* "ghmmwrapper/ghmmunittests.py"
190 (("^(.*)def (testNewXML|testMultipleTransitionClasses|testNewXML)"
192 (string-append indent
193 "@unittest.skip(\"Disabled by Guix\")\n"
196 (add-after 'disable-broken-tests 'autogen
198 (zero? (system* "bash" "autogen.sh")))))))
200 `(("python" ,python-2) ; only Python 2 is supported
201 ("libxml2" ,libxml2)))
203 `(("pkg-config" ,pkg-config)
206 ("autoconf" ,autoconf)
207 ("automake" ,automake)
208 ("libtool" ,libtool)))
209 (home-page "http://ghmm.org")
210 (synopsis "Hidden Markov Model library")
212 "The General Hidden Markov Model library (GHMM) is a C library with
213 additional Python bindings implementing a wide range of types of @dfn{Hidden
214 Markov Models} (HMM) and algorithms: discrete, continuous emissions, basic
215 training, HMM clustering, HMM mixtures.")
216 (license license:lgpl2.0+))))
225 "http://micans.org/mcl/src/mcl-"
226 (string-replace-substring version "." "-")
230 "15xlax3z31lsn62vlg94hkm75nm40q4679amnfg13jm8m2bnhy5m"))))
231 (build-system gnu-build-system)
233 `(#:configure-flags (list "--enable-blast")))
236 (home-page "http://micans.org/mcl/")
237 (synopsis "Clustering algorithm for graphs")
239 "The MCL algorithm is short for the @dfn{Markov Cluster Algorithm}, a
240 fast and scalable unsupervised cluster algorithm for graphs (also known as
241 networks) based on simulation of (stochastic) flow in graphs.")
242 ;; In the LICENCE file and web page it says "The software is licensed
243 ;; under the GNU General Public License, version 3.", but in several of
244 ;; the source code files it suggests GPL3 or later.
245 ;; http://listserver.ebi.ac.uk/pipermail/mcl-users/2016/000376.html
246 (license license:gpl3)))
248 (define-public ocaml-mcl
251 (version "12-068oasis4")
256 "https://github.com/fhcrc/mcl/archive/"
258 (file-name (string-append name "-" version ".tar.gz"))
261 "1l5jbhwjpsj38x8b9698hfpkv75h8hn3kj0gihjhn8ym2cwwv110"))))
262 (build-system ocaml-build-system)
265 (modify-phases %standard-phases
266 (add-before 'configure 'patch-paths
268 (substitute* "configure"
269 (("SHELL = /bin/sh") (string-append "SHELL = "(which "sh"))))
270 (substitute* "setup.ml"
272 (string-append "LDFLAGS=-fPIC\"; \"SHELL=" (which "sh"))))
274 (home-page "https://github.com/fhcrc/mcl")
275 (synopsis "OCaml wrappers around MCL")
277 "This package provides OCaml bindings for the MCL graph clustering
279 (license license:gpl3)))
281 (define-public ocaml4.01-mcl
282 (package-with-ocaml4.01 ocaml-mcl))
284 (define-public randomjungle
286 (name "randomjungle")
292 "http://www.imbs-luebeck.de/imbs/sites/default/files/u59/"
293 "randomjungle-" version ".tar_.gz"))
296 "12c8rf30cla71swx2mf4ww9mfd8jbdw5lnxd7dxhyw1ygrvg6y4w"))))
297 (build-system gnu-build-system)
300 (list (string-append "--with-boost="
301 (assoc-ref %build-inputs "boost")))
303 (modify-phases %standard-phases
305 'configure 'set-CXXFLAGS
307 (setenv "CXXFLAGS" "-fpermissive ")
315 `(("gfortran" ,gfortran)
316 ("gfortran:lib" ,gfortran "lib")))
317 ;; Non-portable assembly instructions are used so building fails on
318 ;; platforms other than x86_64 or i686.
319 (supported-systems '("x86_64-linux" "i686-linux"))
320 (home-page "http://www.imbs-luebeck.de/imbs/de/node/227/")
321 (synopsis "Implementation of the Random Forests machine learning method")
323 "Random Jungle is an implementation of Random Forests. It is supposed to
324 analyse high dimensional data. In genetics, it can be used for analysing big
325 Genome Wide Association (GWA) data. Random Forests is a powerful machine
326 learning method. Most interesting features are variable selection, missing
327 value imputation, classifier creation, generalization error estimation and
328 sample proximities between pairs of cases.")
329 (license license:gpl3+)))
331 (define-public shogun
339 "ftp://shogun-toolbox.org/shogun/releases/"
340 (version-major+minor version)
341 "/sources/shogun-" version ".tar.bz2"))
344 "159nlijnb7mnrv9za80wnm1shwvy45hgrqzn51hxy7gw4z6d6fdb"))
345 (modules '((guix build utils)
349 ;; Remove non-free sources and files referencing them
350 (for-each delete-file
351 (find-files "src/shogun/classifier/svm/"
352 "SVMLight\\.(cpp|h)"))
353 (for-each delete-file
354 (find-files "examples/undocumented/libshogun/"
356 "(classifier_.*svmlight.*|"
357 "evaluation_cross_validation_locked_comparison).cpp")))
358 ;; Remove non-free functions.
359 (define (delete-ifdefs file)
360 (with-atomic-file-replacement file
362 (let loop ((line (read-line in 'concat))
364 (if (eof-object? line)
369 "#endif //USE_SVMLIGHT" line)))
371 "#ifdef USE_SVMLIGHT" line))))
372 (when (or (not skipping?)
373 (and skipping? (not skip-next?)))
375 (loop (read-line in 'concat) skip-next?)))))))
376 (for-each delete-ifdefs (find-files "src/shogun/kernel/"
377 "^Kernel\\.(cpp|h)"))))))
378 (build-system cmake-build-system)
380 '(#:tests? #f ;no check target
382 (modify-phases %standard-phases
383 (add-after 'unpack 'delete-broken-symlinks
385 (for-each delete-file '("applications/arts/data"
386 "applications/asp/data"
387 "applications/easysvm/data"
388 "applications/msplicer/data"
389 "applications/ocr/data"
390 "examples/documented/data"
391 "examples/documented/matlab_static"
392 "examples/documented/octave_static"
393 "examples/undocumented/data"
394 "examples/undocumented/matlab_static"
395 "examples/undocumented/octave_static"
396 "tests/integration/data"
397 "tests/integration/matlab_static"
398 "tests/integration/octave_static"
399 "tests/integration/python_modular/tests"))
401 (add-after 'unpack 'change-R-target-path
402 (lambda* (#:key outputs #:allow-other-keys)
403 (substitute* '("src/interfaces/r_modular/CMakeLists.txt"
404 "src/interfaces/r_static/CMakeLists.txt"
405 "examples/undocumented/r_modular/CMakeLists.txt")
406 (("\\$\\{R_COMPONENT_LIB_PATH\\}")
407 (string-append (assoc-ref outputs "out")
410 (add-after 'unpack 'fix-octave-modules
411 (lambda* (#:key outputs #:allow-other-keys)
412 (substitute* '("src/interfaces/octave_modular/CMakeLists.txt"
413 "src/interfaces/octave_static/CMakeLists.txt")
414 (("^include_directories\\(\\$\\{OCTAVE_INCLUDE_DIRS\\}")
415 "include_directories(${OCTAVE_INCLUDE_DIRS} ${OCTAVE_INCLUDE_DIRS}/octave"))
417 ;; change target directory
418 (substitute* "src/interfaces/octave_modular/CMakeLists.txt"
419 (("\\$\\{OCTAVE_OCT_LOCAL_API_FILE_DIR\\}")
420 (string-append (assoc-ref outputs "out")
421 "/share/octave/packages")))
423 (add-before 'build 'set-HOME
424 ;; $HOME needs to be set at some point during the build phase
425 (lambda _ (setenv "HOME" "/tmp") #t)))
427 (list "-DCMAKE_BUILD_WITH_INSTALL_RPATH=TRUE"
428 "-DUSE_SVMLIGHT=OFF" ;disable proprietary SVMLIGHT
429 ;;"-DJavaModular=ON" ;requires unpackaged jblas
430 ;;"-DRubyModular=ON" ;requires unpackaged ruby-narray
431 ;;"-DPerlModular=ON" ;"FindPerlLibs" does not exist
432 ;;"-DLuaModular=ON" ;fails because lua doesn't build pkgconfig file
439 "-DCmdLineStatic=ON")))
442 ("numpy" ,python-numpy)
443 ("r-minimal" ,r-minimal)
448 ("arpack" ,arpack-ng)
455 `(("pkg-config" ,pkg-config)))
456 ;; Non-portable SSE instructions are used so building fails on platforms
457 ;; other than x86_64.
458 (supported-systems '("x86_64-linux"))
459 (home-page "http://shogun-toolbox.org/")
460 (synopsis "Machine learning toolbox")
462 "The Shogun Machine learning toolbox provides a wide range of unified and
463 efficient Machine Learning (ML) methods. The toolbox seamlessly allows to
464 combine multiple data representations, algorithm classes, and general purpose
465 tools. This enables both rapid prototyping of data pipelines and extensibility
466 in terms of new algorithms.")
467 (license license:gpl3+)))
469 (define-public r-adaptivesparsity
471 (name "r-adaptivesparsity")
475 (uri (cran-uri "AdaptiveSparsity" version))
478 "1az7isvalf3kmdiycrfl6s9k9xqk22k1mc6rh8v0jmcz402qyq8z"))))
480 `((upstream-name . "AdaptiveSparsity")))
481 (build-system r-build-system)
484 (modify-phases %standard-phases
485 (add-after 'unpack 'link-against-armadillo
487 (substitute* "src/Makevars"
488 (("PKG_LIBS=" prefix)
489 (string-append prefix "-larmadillo"))))))))
492 ("r-rcpparmadillo" ,r-rcpparmadillo)))
494 `(("armadillo" ,armadillo)))
495 (home-page "http://cran.r-project.org/web/packages/AdaptiveSparsity")
496 (synopsis "Adaptive sparsity models")
498 "This package implements the Figueiredo machine learning algorithm for
499 adaptive sparsity and the Wong algorithm for adaptively sparse gaussian
501 (license license:lgpl3+)))
503 (define-public r-kernlab
510 (uri (cran-uri "kernlab" version))
513 "0qnaq9x3j2xc6jrmmd98wc6hkzch487s4p3a9lnc00xvahkhgpmr"))))
514 (build-system r-build-system)
515 (home-page "http://cran.r-project.org/web/packages/kernlab")
516 (synopsis "Kernel-based machine learning tools")
518 "This package provides kernel-based machine learning methods for
519 classification, regression, clustering, novelty detection, quantile regression
520 and dimensionality reduction. Among other methods @code{kernlab} includes
521 Support Vector Machines, Spectral Clustering, Kernel PCA, Gaussian Processes
523 (license license:gpl2)))
532 "http://dlib.net/files/dlib-" version ".tar.bz2"))
535 "1mljz02kwkrbggyncxv5fpnyjdybw2qihaacb3js8yfkw12vwpc2"))
536 (modules '((guix build utils)))
539 ;; Delete ~13MB of bundled dependencies.
540 (delete-file-recursively "dlib/external")
541 (delete-file-recursively "docs/dlib/external")
543 (build-system cmake-build-system)
545 ;; Recent releases defaults to "lib64" on 64bit.
546 `(#:configure-flags (list (string-append "-DCMAKE_INSTALL_LIBDIR="
547 (assoc-ref %outputs "out")
550 (modify-phases %standard-phases
551 (add-after 'unpack 'disable-asserts
553 ;; config.h recommends explicitly enabling or disabling asserts
554 ;; when building as a shared library. By default neither is set.
555 (substitute* "dlib/config.h"
556 (("^//#define DLIB_DISABLE_ASSERTS") "#define DLIB_DISABLE_ASSERTS"))
558 (add-after 'disable-asserts 'disable-failing-tests
560 ;; One test times out on MIPS, so we need to disable it.
561 ;; Others are flaky on some platforms.
562 (let* ((system ,(or (%current-target-system)
564 (disabled-tests (cond
565 ((string-prefix? "mips64" system)
566 '("object_detector" ; timeout
568 ((string-prefix? "armhf" system)
569 '("learning_to_track"))
570 ((string-prefix? "i686" system)
575 (substitute* "dlib/test/makefile"
576 (((string-append "SRC \\+= " test "\\.cpp")) "")))
581 ;; No test target, so we build and run the unit tests here.
582 (let ((test-dir (string-append "../dlib-" ,version "/dlib/test")))
583 (with-directory-excursion test-dir
584 (and (zero? (system* "make" "-j" (number->string (parallel-job-count))))
585 (zero? (system* "./dtest" "--runall")))))))
586 (add-after 'install 'delete-static-library
587 (lambda* (#:key outputs #:allow-other-keys)
588 (delete-file (string-append (assoc-ref outputs "out")
592 `(("pkg-config" ,pkg-config)))
599 ("openblas" ,openblas)
602 "Toolkit for making machine learning and data analysis applications in C++")
604 "Dlib is a modern C++ toolkit containing machine learning algorithms and
605 tools. It is used in both industry and academia in a wide range of domains
606 including robotics, embedded devices, mobile phones, and large high performance
607 computing environments.")
608 (home-page "http://dlib.net")
609 (license license:boost1.0)))
611 (define-public python-scikit-learn
613 (name "python-scikit-learn")
619 "https://github.com/scikit-learn/scikit-learn/archive/"
621 (file-name (string-append name "-" version ".tar.gz"))
624 "18n8775kyfwbvcjjjzda9c5sqy4737c0hrmj6qj1ps2jmlqzair9"))))
625 (build-system python-build-system)
628 (modify-phases %standard-phases
630 (add-after 'install 'check
631 ;; Running tests from the source directory requires
632 ;; an "inplace" build with paths relative to CWD.
633 ;; http://scikit-learn.org/stable/developers/advanced_installation.html#testing
634 ;; Use the installed version instead.
635 (lambda* (#:key inputs outputs #:allow-other-keys)
636 (add-installed-pythonpath inputs outputs)
637 ;; some tests require access to "$HOME"
638 (setenv "HOME" "/tmp")
639 ;; Step out of the source directory just to be sure.
641 (zero? (system* "nosetests" "-v" "sklearn")))))))
643 `(("openblas" ,openblas)))
645 `(("python-nose" ,python-nose)
646 ("python-cython" ,python-cython)))
648 `(("python-numpy" ,python-numpy)
649 ("python-scipy" ,python-scipy)))
650 (home-page "http://scikit-learn.org/")
651 (synopsis "Machine Learning in Python")
653 "Scikit-learn provides simple and efficient tools for data
654 mining and data analysis.")
655 (license license:bsd-3)))
657 (define-public python2-scikit-learn
658 (package-with-python2 python-scikit-learn))