1 ;;; GNU Guix --- Functional package management for GNU
2 ;;; Copyright © 2015, 2016, 2017, 2018 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>
7 ;;; Copyright © 2018 Kei Kebreau <kkebreau@posteo.net>
8 ;;; Copyright © 2018 Mark Meyer <mark@ofosos.org>
9 ;;; Copyright © 2018 Ben Woodcroft <donttrustben@gmail.com>
10 ;;; Copyright © 2018 Fis Trivial <ybbs.daans@hotmail.com>
11 ;;; Copyright © 2018 Julien Lepiller <julien@lepiller.eu>
12 ;;; Copyright © 2018 Björn Höfling <bjoern.hoefling@bjoernhoefling.de>
14 ;;; This file is part of GNU Guix.
16 ;;; GNU Guix is free software; you can redistribute it and/or modify it
17 ;;; under the terms of the GNU General Public License as published by
18 ;;; the Free Software Foundation; either version 3 of the License, or (at
19 ;;; your option) any later version.
21 ;;; GNU Guix is distributed in the hope that it will be useful, but
22 ;;; WITHOUT ANY WARRANTY; without even the implied warranty of
23 ;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24 ;;; GNU General Public License for more details.
26 ;;; You should have received a copy of the GNU General Public License
27 ;;; along with GNU Guix. If not, see <http://www.gnu.org/licenses/>.
29 (define-module (gnu packages machine-learning)
30 #:use-module ((guix licenses) #:prefix license:)
31 #:use-module (guix packages)
32 #:use-module (guix utils)
33 #:use-module (guix download)
34 #:use-module (guix svn-download)
35 #:use-module (guix build-system cmake)
36 #:use-module (guix build-system gnu)
37 #:use-module (guix build-system ocaml)
38 #:use-module (guix build-system python)
39 #:use-module (guix build-system r)
40 #:use-module (guix git-download)
41 #:use-module (gnu packages)
42 #:use-module (gnu packages algebra)
43 #:use-module (gnu packages autotools)
44 #:use-module (gnu packages boost)
45 #:use-module (gnu packages check)
46 #:use-module (gnu packages compression)
47 #:use-module (gnu packages cran)
48 #:use-module (gnu packages dejagnu)
49 #:use-module (gnu packages gcc)
50 #:use-module (gnu packages image)
51 #:use-module (gnu packages maths)
52 #:use-module (gnu packages mpi)
53 #:use-module (gnu packages ocaml)
54 #:use-module (gnu packages onc-rpc)
55 #:use-module (gnu packages perl)
56 #:use-module (gnu packages pkg-config)
57 #:use-module (gnu packages python)
58 #:use-module (gnu packages python-xyz)
59 #:use-module (gnu packages statistics)
60 #:use-module (gnu packages swig)
61 #:use-module (gnu packages xml)
62 #:use-module (gnu packages xorg))
65 ;; The last release is >100 commits behind, so we package from git.
66 (let ((commit "d71d54788bee56ba4cf7522801270152da5209d7"))
69 (version (string-append "2.2.0-1." (string-take commit 8)))
73 (url "https://github.com/libfann/fann.git")
75 (file-name (string-append name "-" version "-checkout"))
78 "0ibwpfrjs6q2lijs8slxjgzb2llcl6rk3v2ski4r6215g5jjhg3x"))))
79 (build-system cmake-build-system)
82 (modify-phases %standard-phases
84 (lambda* (#:key outputs #:allow-other-keys)
85 (let* ((out (assoc-ref outputs "out")))
86 (with-directory-excursion (string-append (getcwd) "/tests")
87 (invoke "./fann_tests"))))))))
88 (home-page "http://leenissen.dk/fann/wp/")
89 (synopsis "Fast Artificial Neural Network")
91 "FANN is a free open source neural network library, which implements
92 multilayer artificial neural networks in C with support for both fully
93 connected and sparsely connected networks.")
94 (license license:lgpl2.1))))
103 (uri (string-append "https://www.csie.ntu.edu.tw/~cjlin/libsvm/"
104 name "-" version ".tar.gz"))
107 "0zd7s19y5vb7agczl6456bn45cj1y64739sslaskw1qk7dywd0bd"))))
108 (build-system gnu-build-system)
110 `(#:tests? #f ;no "check" target
111 #:phases (modify-phases %standard-phases
114 'install ; no ‘install’ target
115 (lambda* (#:key outputs #:allow-other-keys)
116 (let* ((out (assoc-ref outputs "out"))
117 (bin (string-append out "/bin/")))
119 (for-each (lambda (file)
120 (copy-file file (string-append bin file)))
125 (home-page "http://www.csie.ntu.edu.tw/~cjlin/libsvm/")
126 (synopsis "Library for Support Vector Machines")
128 "LIBSVM is a machine learning library for support vector
129 classification, (C-SVC, nu-SVC), regression (epsilon-SVR, nu-SVR) and
130 distribution estimation (one-class SVM). It supports multi-class
132 (license license:bsd-3)))
134 (define-public python-libsvm
135 (package (inherit libsvm)
136 (name "python-libsvm")
137 (build-system gnu-build-system)
139 `(#:tests? #f ;no "check" target
140 #:make-flags '("-C" "python")
142 (modify-phases %standard-phases
145 'install ; no ‘install’ target
146 (lambda* (#:key inputs outputs #:allow-other-keys)
147 (let ((site (string-append (assoc-ref outputs "out")
151 (assoc-ref inputs "python") 5) 3)
153 (substitute* "python/svm.py"
154 (("../libsvm.so.2") "libsvm.so.2"))
156 (for-each (lambda (file)
157 (copy-file file (string-append site (basename file))))
158 (find-files "python" "\\.py"))
159 (copy-file "libsvm.so.2"
160 (string-append site "libsvm.so.2")))
163 `(("python" ,python)))
164 (synopsis "Python bindings of libSVM")))
167 ;; The latest release candidate is several years and a couple of fixes have
168 ;; been published since. This is why we download the sources from the SVN
170 (let ((svn-revision 2341))
173 (version (string-append "0.9-rc3-0." (number->string svn-revision)))
177 (url "http://svn.code.sf.net/p/ghmm/code/trunk")
178 (revision svn-revision)))
179 (file-name (string-append name "-" version))
182 "0qbq1rqp94l530f043qzp8aw5lj7dng9wq0miffd7spd1ff638wq"))))
183 (build-system gnu-build-system)
185 `(#:imported-modules (,@%gnu-build-system-modules
186 (guix build python-build-system))
188 (modify-phases %standard-phases
189 (add-after 'unpack 'enter-dir
190 (lambda _ (chdir "ghmm") #t))
192 (add-after 'install 'check
193 (assoc-ref %standard-phases 'check))
194 (add-before 'check 'fix-PYTHONPATH
195 (lambda* (#:key inputs outputs #:allow-other-keys)
196 (let ((python-version ((@@ (guix build python-build-system)
198 (assoc-ref inputs "python"))))
200 (string-append (getenv "PYTHONPATH")
201 ":" (assoc-ref outputs "out")
202 "/lib/python" python-version
205 (add-after 'enter-dir 'fix-runpath
206 (lambda* (#:key outputs #:allow-other-keys)
207 (substitute* "ghmmwrapper/setup.py"
208 (("^(.*)extra_compile_args = \\[" line indent)
209 (string-append indent
210 "extra_link_args = [\"-Wl,-rpath="
211 (assoc-ref outputs "out") "/lib\"],\n"
214 (assoc-ref outputs "out")
217 (add-after 'enter-dir 'disable-broken-tests
219 (substitute* "tests/Makefile.am"
220 ;; GHMM_SILENT_TESTS is assumed to be a command.
221 (("TESTS_ENVIRONMENT.*") "")
222 ;; Do not build broken tests.
226 (("label_higher_order_test.*$")
227 "label_higher_order_test\n"))
229 ;; These Python unittests are broken as there is no gato.
230 ;; See https://sourceforge.net/p/ghmm/support-requests/3/
231 (substitute* "ghmmwrapper/ghmmunittests.py"
232 (("^(.*)def (testNewXML|testMultipleTransitionClasses|testNewXML)"
234 (string-append indent
235 "@unittest.skip(\"Disabled by Guix\")\n"
238 (add-after 'disable-broken-tests 'autogen
240 (invoke "bash" "autogen.sh"))))))
242 `(("python" ,python-2) ; only Python 2 is supported
243 ("libxml2" ,libxml2)))
245 `(("pkg-config" ,pkg-config)
248 ("autoconf" ,autoconf)
249 ("automake" ,automake)
250 ("libtool" ,libtool)))
251 (home-page "http://ghmm.org")
252 (synopsis "Hidden Markov Model library")
254 "The General Hidden Markov Model library (GHMM) is a C library with
255 additional Python bindings implementing a wide range of types of @dfn{Hidden
256 Markov Models} (HMM) and algorithms: discrete, continuous emissions, basic
257 training, HMM clustering, HMM mixtures.")
258 (license license:lgpl2.0+))))
267 "http://micans.org/mcl/src/mcl-"
268 (string-replace-substring version "." "-")
272 "15xlax3z31lsn62vlg94hkm75nm40q4679amnfg13jm8m2bnhy5m"))))
273 (build-system gnu-build-system)
275 `(#:configure-flags (list "--enable-blast")))
278 (home-page "http://micans.org/mcl/")
279 (synopsis "Clustering algorithm for graphs")
281 "The MCL algorithm is short for the @dfn{Markov Cluster Algorithm}, a
282 fast and scalable unsupervised cluster algorithm for graphs (also known as
283 networks) based on simulation of (stochastic) flow in graphs.")
284 ;; In the LICENCE file and web page it says "The software is licensed
285 ;; under the GNU General Public License, version 3.", but in several of
286 ;; the source code files it suggests GPL3 or later.
287 ;; http://listserver.ebi.ac.uk/pipermail/mcl-users/2016/000376.html
288 (license license:gpl3)))
290 (define-public ocaml-mcl
293 (version "12-068oasis4")
298 "https://github.com/fhcrc/mcl/archive/"
300 (file-name (string-append name "-" version ".tar.gz"))
303 "1l5jbhwjpsj38x8b9698hfpkv75h8hn3kj0gihjhn8ym2cwwv110"))))
304 (build-system ocaml-build-system)
306 `(#:ocaml ,ocaml-4.02
307 #:findlib ,ocaml4.02-findlib
309 (modify-phases %standard-phases
310 (add-before 'configure 'patch-paths
312 (substitute* "configure"
313 (("SHELL = /bin/sh") (string-append "SHELL = "(which "sh"))))
314 (substitute* "setup.ml"
316 (string-append "LDFLAGS=-fPIC\"; \"SHELL=" (which "sh"))))
318 (home-page "https://github.com/fhcrc/mcl")
319 (synopsis "OCaml wrappers around MCL")
321 "This package provides OCaml bindings for the MCL graph clustering
323 (license license:gpl3)))
325 (define-public ocaml4.01-mcl
326 (package-with-ocaml4.01 ocaml-mcl))
328 (define-public randomjungle
330 (name "randomjungle")
336 "https://www.imbs.uni-luebeck.de/fileadmin/files/Software"
337 "/randomjungle/randomjungle-" version ".tar_.gz"))
338 (patches (search-patches "randomjungle-disable-static-build.patch"))
341 "12c8rf30cla71swx2mf4ww9mfd8jbdw5lnxd7dxhyw1ygrvg6y4w"))))
342 (build-system gnu-build-system)
345 (list "--disable-static"
346 (string-append "--with-boost="
347 (assoc-ref %build-inputs "boost")))
349 (modify-phases %standard-phases
351 'configure 'set-CXXFLAGS
353 (setenv "CXXFLAGS" "-fpermissive ")
361 `(("gfortran" ,gfortran)
362 ("gfortran:lib" ,gfortran "lib")))
363 ;; Non-portable assembly instructions are used so building fails on
364 ;; platforms other than x86_64 or i686.
365 (supported-systems '("x86_64-linux" "i686-linux"))
366 (home-page "https://www.imbs.uni-luebeck.de/forschung/software/details.html#c224")
367 (synopsis "Implementation of the Random Forests machine learning method")
369 "Random Jungle is an implementation of Random Forests. It is supposed to
370 analyse high dimensional data. In genetics, it can be used for analysing big
371 Genome Wide Association (GWA) data. Random Forests is a powerful machine
372 learning method. Most interesting features are variable selection, missing
373 value imputation, classifier creation, generalization error estimation and
374 sample proximities between pairs of cases.")
375 (license license:gpl3+)))
377 (define-public shogun
385 "ftp://shogun-toolbox.org/shogun/releases/"
386 (version-major+minor version)
387 "/sources/shogun-" version ".tar.bz2"))
390 "1rn9skm3nw6hr7mr3lgp2gfqhi7ii0lyxck7qmqnf8avq349s5jp"))
391 (modules '((guix build utils)
395 ;; Remove non-free sources and files referencing them
396 (for-each delete-file
397 (find-files "src/shogun/classifier/svm/"
398 "SVMLight\\.(cpp|h)"))
399 (for-each delete-file
400 (find-files "examples/undocumented/libshogun/"
402 "(classifier_.*svmlight.*|"
403 "evaluation_cross_validation_locked_comparison).cpp")))
404 ;; Remove non-free functions.
405 (define (delete-ifdefs file)
406 (with-atomic-file-replacement file
408 (let loop ((line (read-line in 'concat))
410 (if (eof-object? line)
415 "#endif //USE_SVMLIGHT" line)))
417 "#ifdef USE_SVMLIGHT" line))))
418 (when (or (not skipping?)
419 (and skipping? (not skip-next?)))
421 (loop (read-line in 'concat) skip-next?)))))))
422 (for-each delete-ifdefs
424 (find-files "src/shogun/classifier/mkl"
425 "^MKLClassification\\.cpp")
426 (find-files "src/shogun/classifier/svm"
427 "^SVMLightOneClass\\.(cpp|h)")
428 (find-files "src/shogun/multiclass"
429 "^ScatterSVM\\.(cpp|h)")
430 (find-files "src/shogun/kernel/"
431 "^(Kernel|CombinedKernel|ProductKernel)\\.(cpp|h)")
432 (find-files "src/shogun/regression/svr"
433 "^(MKLRegression|SVRLight)\\.(cpp|h)")
434 (find-files "src/shogun/transfer/domain_adaptation"
435 "^DomainAdaptationSVM\\.(cpp|h)")))
437 (build-system cmake-build-system)
439 '(#:tests? #f ;no check target
441 (modify-phases %standard-phases
442 (add-after 'unpack 'delete-broken-symlinks
444 (for-each delete-file '("applications/arts/data"
445 "applications/asp/data"
446 "applications/easysvm/data"
447 "applications/msplicer/data"
448 "applications/ocr/data"
450 "examples/undocumented/data"))
452 (add-after 'unpack 'change-R-target-path
453 (lambda* (#:key outputs #:allow-other-keys)
454 (substitute* '("src/interfaces/r/CMakeLists.txt"
455 "examples/meta/r/CMakeLists.txt")
456 (("\\$\\{R_COMPONENT_LIB_PATH\\}")
457 (string-append (assoc-ref outputs "out")
460 (add-after 'unpack 'fix-octave-modules
461 (lambda* (#:key outputs #:allow-other-keys)
462 (substitute* "src/interfaces/octave/CMakeLists.txt"
463 (("^include_directories\\(\\$\\{OCTAVE_INCLUDE_DIRS\\}")
464 "include_directories(${OCTAVE_INCLUDE_DIRS} ${OCTAVE_INCLUDE_DIRS}/octave")
465 ;; change target directory
466 (("\\$\\{OCTAVE_OCT_LOCAL_API_FILE_DIR\\}")
467 (string-append (assoc-ref outputs "out")
468 "/share/octave/packages")))
469 (substitute* '("src/interfaces/octave/swig_typemaps.i"
470 "src/interfaces/octave/sg_print_functions.cpp")
471 ;; "octave/config.h" and "octave/oct-obj.h" deprecated in Octave.
472 (("octave/config\\.h") "octave/octave-config.h")
473 (("octave/oct-obj.h") "octave/ovl.h"))
475 (add-after 'unpack 'move-rxcpp
476 (lambda* (#:key inputs #:allow-other-keys)
477 (let ((rxcpp-dir "shogun/third-party/rxcpp"))
479 (install-file (assoc-ref inputs "rxcpp") rxcpp-dir)
481 (add-before 'build 'set-HOME
482 ;; $HOME needs to be set at some point during the build phase
483 (lambda _ (setenv "HOME" "/tmp") #t)))
485 (list "-DCMAKE_BUILD_WITH_INSTALL_RPATH=TRUE"
486 "-DUSE_SVMLIGHT=OFF" ;disable proprietary SVMLIGHT
487 "-DBUILD_META_EXAMPLES=OFF" ;requires unpackaged ctags
488 ;;"-DINTERFACE_JAVA=ON" ;requires unpackaged jblas
489 ;;"-DINTERFACE_RUBY=ON" ;requires unpackaged ruby-narray
490 ;;"-DINTERFACE_PERL=ON" ;"FindPerlLibs" does not exist
491 ;;"-DINTERFACE_LUA=ON" ;fails because lua doesn't build pkgconfig file
492 "-DINTERFACE_OCTAVE=ON"
493 "-DINTERFACE_PYTHON=ON"
494 "-DINTERFACE_R=ON")))
497 ("numpy" ,python-numpy)
498 ("r-minimal" ,r-minimal)
499 ("octave" ,octave-cli)
504 ("arpack" ,arpack-ng)
511 `(("pkg-config" ,pkg-config)
513 ;; Non-portable SSE instructions are used so building fails on platforms
514 ;; other than x86_64.
515 (supported-systems '("x86_64-linux"))
516 (home-page "http://shogun-toolbox.org/")
517 (synopsis "Machine learning toolbox")
519 "The Shogun Machine learning toolbox provides a wide range of unified and
520 efficient Machine Learning (ML) methods. The toolbox seamlessly allows to
521 combine multiple data representations, algorithm classes, and general purpose
522 tools. This enables both rapid prototyping of data pipelines and extensibility
523 in terms of new algorithms.")
524 (license license:gpl3+)))
533 (uri (string-append "https://github.com/ReactiveX/RxCpp/archive/v"
537 "0y2isr8dy2n1yjr9c5570kpc9lvdlch6jv0jvw000amwn5d3krsh"))
538 (file-name (string-append name "-" version ".tar.gz"))))
539 (build-system cmake-build-system)
542 (modify-phases %standard-phases
543 (add-after 'unpack 'remove-werror
545 (substitute* (find-files ".")
550 (invoke "ctest"))))))
552 `(("catch" ,catch-framework)))
553 (home-page "http://reactivex.io/")
554 (synopsis "Reactive Extensions for C++")
556 "The Reactive Extensions for C++ (RxCpp) is a library of algorithms for
557 values-distributed-in-time. ReactiveX is a library for composing asynchronous
558 and event-based programs by using observable sequences.
560 It extends the observer pattern to support sequences of data and/or events and
561 adds operators that allow you to compose sequences together declaratively while
562 abstracting away concerns about things like low-level threading,
563 synchronization, thread-safety, concurrent data structures, and non-blocking
565 (license license:asl2.0)))
567 (define-public r-adaptivesparsity
569 (name "r-adaptivesparsity")
573 (uri (cran-uri "AdaptiveSparsity" version))
576 "0imr5m8mll9j6n4icsv6z9rl5kbnwsp9wvzrg7n90nnmcxq2cz91"))))
578 `((upstream-name . "AdaptiveSparsity")))
579 (build-system r-build-system)
582 (modify-phases %standard-phases
583 (add-after 'unpack 'link-against-armadillo
585 (substitute* "src/Makevars"
586 (("PKG_LIBS=" prefix)
587 (string-append prefix "-larmadillo"))))))))
590 ("r-matrix" ,r-matrix)
592 ("r-rcpparmadillo" ,r-rcpparmadillo)))
594 `(("armadillo" ,armadillo)))
595 (home-page "https://cran.r-project.org/web/packages/AdaptiveSparsity")
596 (synopsis "Adaptive sparsity models")
598 "This package implements the Figueiredo machine learning algorithm for
599 adaptive sparsity and the Wong algorithm for adaptively sparse gaussian
601 (license license:lgpl3+)))
603 (define-public r-kernlab
610 (uri (cran-uri "kernlab" version))
613 "1m0xqf6gyvwayz7w3c83y32ayvnlz0jicj8ijk808zq9sh7dbbgn"))))
614 (build-system r-build-system)
615 (home-page "https://cran.r-project.org/web/packages/kernlab")
616 (synopsis "Kernel-based machine learning tools")
618 "This package provides kernel-based machine learning methods for
619 classification, regression, clustering, novelty detection, quantile regression
620 and dimensionality reduction. Among other methods @code{kernlab} includes
621 Support Vector Machines, Spectral Clustering, Kernel PCA, Gaussian Processes
623 (license license:gpl2)))
632 "http://dlib.net/files/dlib-" version ".tar.bz2"))
635 "1mljz02kwkrbggyncxv5fpnyjdybw2qihaacb3js8yfkw12vwpc2"))
636 (modules '((guix build utils)))
639 ;; Delete ~13MB of bundled dependencies.
640 (delete-file-recursively "dlib/external")
641 (delete-file-recursively "docs/dlib/external")
643 (build-system cmake-build-system)
646 (modify-phases %standard-phases
647 (add-after 'unpack 'disable-asserts
649 ;; config.h recommends explicitly enabling or disabling asserts
650 ;; when building as a shared library. By default neither is set.
651 (substitute* "dlib/config.h"
652 (("^//#define DLIB_DISABLE_ASSERTS") "#define DLIB_DISABLE_ASSERTS"))
654 (add-after 'disable-asserts 'disable-failing-tests
656 ;; One test times out on MIPS, so we need to disable it.
657 ;; Others are flaky on some platforms.
658 (let* ((system ,(or (%current-target-system)
660 (disabled-tests (cond
661 ((string-prefix? "mips64" system)
662 '("object_detector" ; timeout
664 ((string-prefix? "armhf" system)
665 '("learning_to_track"))
666 ((string-prefix? "i686" system)
671 (substitute* "dlib/test/makefile"
672 (((string-append "SRC \\+= " test "\\.cpp")) "")))
677 ;; No test target, so we build and run the unit tests here.
678 (let ((test-dir (string-append "../dlib-" ,version "/dlib/test")))
679 (with-directory-excursion test-dir
680 (invoke "make" "-j" (number->string (parallel-job-count)))
681 (invoke "./dtest" "--runall"))
683 (add-after 'install 'delete-static-library
684 (lambda* (#:key outputs #:allow-other-keys)
685 (delete-file (string-append (assoc-ref outputs "out")
689 `(("pkg-config" ,pkg-config)
698 ("openblas" ,openblas)
701 "Toolkit for making machine learning and data analysis applications in C++")
703 "Dlib is a modern C++ toolkit containing machine learning algorithms and
704 tools. It is used in both industry and academia in a wide range of domains
705 including robotics, embedded devices, mobile phones, and large high performance
706 computing environments.")
707 (home-page "http://dlib.net")
708 (license license:boost1.0)))
710 (define-public python-scikit-learn
712 (name "python-scikit-learn")
718 (url "https://github.com/scikit-learn/scikit-learn.git")
720 (file-name (git-file-name name version))
723 "0fkhwg3xn1s7ln9q1szq6kwc4jhwvjh8w4kmv9wcrqy7cq3lbv0d"))))
724 (build-system python-build-system)
727 (modify-phases %standard-phases
728 (add-after 'build 'build-ext
729 (lambda _ (invoke "python" "setup.py" "build_ext" "--inplace") #t))
732 ;; Restrict OpenBLAS threads to prevent segfaults while testing!
733 (setenv "OPENBLAS_NUM_THREADS" "1")
735 ;; Some tests require write access to $HOME.
736 (setenv "HOME" "/tmp")
738 (invoke "pytest" "sklearn" "-m" "not network")))
739 ;; FIXME: This fails with permission denied
740 (delete 'reset-gzip-timestamps))))
742 `(("openblas" ,openblas)))
744 `(("python-pytest" ,python-pytest)
745 ("python-pandas" ,python-pandas) ;for tests
746 ("python-cython" ,python-cython)))
748 `(("python-numpy" ,python-numpy)
749 ("python-scipy" ,python-scipy)))
750 (home-page "http://scikit-learn.org/")
751 (synopsis "Machine Learning in Python")
753 "Scikit-learn provides simple and efficient tools for data mining and
755 (license license:bsd-3)))
757 (define-public python2-scikit-learn
758 (package-with-python2 python-scikit-learn))
760 (define-public python-autograd
761 (let* ((commit "442205dfefe407beffb33550846434baa90c4de7")
763 (version (git-version "0.0.0" revision commit)))
765 (name "python-autograd")
766 (home-page "https://github.com/HIPS/autograd")
774 "189sv2xb0mwnjawa9z7mrgdglc1miaq93pnck26r28fi1jdwg0z4"))
775 (file-name (git-file-name name version))))
777 (build-system python-build-system)
779 `(("python-nose" ,python-nose)
780 ("python-pytest" ,python-pytest)))
782 `(("python-future" ,python-future)
783 ("python-numpy" ,python-numpy)))
785 `(#:phases (modify-phases %standard-phases
788 (invoke "py.test" "-v"))))))
789 (synopsis "Efficiently computes derivatives of NumPy code")
790 (description "Autograd can automatically differentiate native Python and
791 NumPy code. It can handle a large subset of Python's features, including loops,
792 ifs, recursion and closures, and it can even take derivatives of derivatives
793 of derivatives. It supports reverse-mode differentiation
794 (a.k.a. backpropagation), which means it can efficiently take gradients of
795 scalar-valued functions with respect to array-valued arguments, as well as
796 forward-mode differentiation, and the two can be composed arbitrarily. The
797 main intended application of Autograd is gradient-based optimization.")
798 (license license:expat))))
800 (define-public python2-autograd
801 (package-with-python2 python-autograd))
803 (define-public lightgbm
810 "https://github.com/Microsoft/LightGBM/archive/v"
814 "132zf0yk0545mg72hyzxm102g3hpb6ixx9hnf8zd2k55gas6cjj1"))
815 (file-name (string-append name "-" version ".tar.gz"))))
817 `(("python-pytest" ,python-pytest)
818 ("python-nose" ,python-nose)))
820 `(("openmpi" ,openmpi)))
822 `(("python-numpy" ,python-numpy)
823 ("python-scipy" ,python-scipy)))
828 (modify-phases %standard-phases
830 (lambda* (#:key outputs #:allow-other-keys)
831 (with-directory-excursion ,(string-append "../LightGBM-" version)
832 (invoke "pytest" "tests/c_api_test/test_.py")))))))
833 (build-system cmake-build-system)
834 (home-page "https://github.com/Microsoft/LightGBM")
835 (synopsis "Gradient boosting framework based on decision tree algorithms")
836 (description "LightGBM is a gradient boosting framework that uses tree
837 based learning algorithms. It is designed to be distributed and efficient with
838 the following advantages:
841 @item Faster training speed and higher efficiency
842 @item Lower memory usage
843 @item Better accuracy
844 @item Parallel and GPU learning supported (not enabled in this package)
845 @item Capable of handling large-scale data
847 (license license:expat)))
849 (define-public vowpal-wabbit
850 ;; Language bindings not included.
852 (name "vowpal-wabbit")
857 "https://github.com/JohnLangford/vowpal_wabbit/archive/"
861 "0clp2kb7rk5sckhllxjr5a651awf4s8dgzg4659yh4hf5cqnf0gr"))
862 (file-name (string-append name "-" version ".tar.gz"))))
868 (list (string-append "--with-boost="
869 (assoc-ref %build-inputs "boost")))))
870 (build-system gnu-build-system)
871 (home-page "https://github.com/JohnLangford/vowpal_wabbit")
872 (synopsis "Fast machine learning library for online learning")
873 (description "Vowpal Wabbit is a machine learning system with techniques
874 such as online, hashing, allreduce, reductions, learning2search, active, and
875 interactive learning.")
876 (license license:bsd-3)))
878 (define-public python2-fastlmm
880 (name "python2-fastlmm")
885 (uri (pypi-uri "fastlmm" version ".zip"))
888 "1q8c34rpmwkfy3r4d5172pzdkpfryj561897z9r3x22gq7813x1m"))))
889 (build-system python-build-system)
891 `(#:python ,python-2)) ; only Python 2.7 is supported
893 `(("python2-numpy" ,python2-numpy)
894 ("python2-scipy" ,python2-scipy)
895 ("python2-matplotlib" ,python2-matplotlib)
896 ("python2-pandas" ,python2-pandas)
897 ("python2-scikit-learn" ,python2-scikit-learn)
898 ("python2-pysnptools" ,python2-pysnptools)))
901 ("python2-cython" ,python2-cython)
902 ("python2-mock" ,python2-mock)
903 ("python2-nose" ,python2-nose)))
904 (home-page "http://research.microsoft.com/en-us/um/redmond/projects/mscompbio/fastlmm/")
905 (synopsis "Perform genome-wide association studies on large data sets")
907 "FaST-LMM, which stands for Factored Spectrally Transformed Linear Mixed
908 Models, is a program for performing both single-SNP and SNP-set genome-wide
909 association studies (GWAS) on extremely large data sets.")
910 (license license:asl2.0)))