Merge branch 'master' into liaad

* master: (28 commits)
  remove obsolete URL
  README.md: add CI status, more compact overview
  fix Travis issues, 3rd attempt (zenogantner#240)
  package repair, version bump
  fix Travis issues, next attempt (zenogantner#240)
  faster CanPredict() methods for the external recommenders
  get rid of Utils.Memoize (zenogantner#231)
  play with Travis-CI settings
  touch target so script is not run every time
  331: README -> README.md
  480 rm SigmoidCombinedAsymmetricFactorModel from load/save integration test
  allow modifying floating point format with --prediction-line
  update (c) line
  rm monster Perl one-liner
  add InvariantCulture to Format() calls
  more verbose test output
  bump version number to 3.13
  remove some left-over .mdb references
  follow-up to zenogantner#423 zenogantner#479: simplify code, also apply change to ExternalRatingPredictor
  Reduced complexity of ExternalItemRecommender
  ...

# Conflicts:
#	src/MyMediaLite-nogui.sln

.travis.yml

@@ -1,7 +1,7 @@
 language: csharp
 mono:
  - latest
- - 3.8.0
+ - beta
 solution: src/MyMediaLite-nogui.sln
 install:
   - nuget restore src/MyMediaLite-nogui.sln

Makefile

@@ -3,7 +3,7 @@ EDITOR:=editor
 GENDARME_OPTIONS:=--quiet --severity critical+
 SRC_DIR:=src
 PREFIX:=/usr/local
-VERSION:=3.12
+VERSION:=3.13
 HOMEPAGE:=../mymedialite.net/public_html
 HOMEPAGE_SRC:=../mymedialite.net/src
 HOMEPAGE_INC:=../mymedialite.net/lib
@@ -58,11 +58,11 @@ MyMediaLite-${VERSION}.tar.gz:
 	mkdir MyMediaLite-${VERSION}/lib/mymedialite
 	cp ${ITEM_REC_DIR}/bin/Debug/*.exe MyMediaLite-${VERSION}/lib/mymedialite
 	cp ${ITEM_REC_DIR}/bin/Debug/*.dll MyMediaLite-${VERSION}/lib/mymedialite
-	cp ${ITEM_REC_DIR}/bin/Debug/*.mdb MyMediaLite-${VERSION}/lib/mymedialite
+	cp ${ITEM_REC_DIR}/bin/Debug/*.pdb MyMediaLite-${VERSION}/lib/mymedialite
 	cp ${RATING_PRED_DIR}/bin/Debug/*.exe MyMediaLite-${VERSION}/lib/mymedialite
-	cp ${RATING_PRED_DIR}/bin/Debug/*.exe.mdb MyMediaLite-${VERSION}/lib/mymedialite
+	cp ${RATING_PRED_DIR}/bin/Debug/*.exe.pdb MyMediaLite-${VERSION}/lib/mymedialite
 	cp ${RATING_RANK_DIR}/bin/Debug/*.exe MyMediaLite-${VERSION}/lib/mymedialite
-	cp ${RATING_RANK_DIR}/bin/Debug/*.exe.mdb MyMediaLite-${VERSION}/lib/mymedialite
+	cp ${RATING_RANK_DIR}/bin/Debug/*.exe.pdb MyMediaLite-${VERSION}/lib/mymedialite
 	tar -cvzf MyMediaLite-${VERSION}.tar.gz MyMediaLite-${VERSION}
 	rm -rf MyMediaLite-${VERSION}
 
@@ -113,6 +113,7 @@ data:
 
 data/ml-100k/u.data: data
 	scripts/download_ml-100k.sh
+	touch $@
 
 download-movielens: data/ml-100k/u.data
 	scripts/download_movielens.sh

README → README.md

@@ -1,71 +1,76 @@
-MyMediaLite - a recommender system algorithm library
+MyMediaLite - a free recommender system algorithm library
+=========================================================
+
+[![Build Status](https://travis-ci.org/zenogantner/MyMediaLite.svg?branch=master)](https://travis-ci.org/zenogantner/MyMediaLite)
+
 
 Features
 --------
+
  - Dozens of different recommendation methods,
-   - methods can use collaborative, attribute/content, and relational data,
+    - methods can use collaborative, attribute/content, and relational data,
  - support for incremental training for most models.
  - Ready to use:
-   - Includes evaluation routines for rating and item prediction;
-     quality measures MAE, NAME, RMSE, CBD, AUC, MAP, precision@N, recall@N,
-     NDCG, MRR; and
-   - command line tools that read a simple text-based input format.
- - Compactness: Core library is about 275 KB "big".
- - Portability: Written in C#, for the .NET platform;
+    - Includes evaluation routines for rating and item prediction;
+      quality measures MAE, NAME, RMSE, CBD, AUC, MAP, precision@N, recall@N,
+      NDCG, MRR; and
+    - command line tools that read a simple text-based input format.
+ - Compact: Core library is about 275 KB "big".
+ - Portable: Written in C#, for the .NET platform;
    runs on every architecture where Mono works: Linux, Windows, Mac OS X.
- - Freedom: Free/Open Source software, distributed under the terms of the
-            GNU General Public License (GPL).
 
 
 Feedback and Contributions
 --------------------------
+
 We are always happy about feedback, and encourage MyMediaLite's users to
 contribute code to the project.
 
-Just fork it on GitHub and send pull requests!
+Just fork it on [GitHub](https://github.com/zenogantner/MyMediaLite) and send pull requests!
 
-http://mymedialite.net/contribute.html
+Bugs and feature requests can be reported on our
+[mailing list](https://groups.google.com/group/mymedialite) or in our
+[issue tracker](https://github.com/zenogantner/MyMediaLite/issues).
 
-Bugs and feature requests can be reported on our mailing list or in our issue
-tracker:
-https://groups.google.com/group/mymedialite
-https://github.com/zenogantner/MyMediaLite/issues
 
 Installation
 ------------
+
 See doc/Installation for installation instructions.
 
+
 Documentation
 -------------
-See doc/ and the website for more documentation.
 
-Website
--------
-  http://mymedialite.net
-  http://twitter.com/mymedialite
+See doc/ and the [website](http://mymedialite.net) for more documentation.
+
 
 Citing MyMediaLite
 ------------------
+
 If you use MyMediaLite for your research, it would be nice to acknowledge it
 in your papers by citing the following paper:
 
 Zeno Gantner, Steffen Rendle, Christoph Freudenthaler, Lars Schmidt-Thieme:
 MyMediaLite: A Free Recommender System Library. RecSys 2011
 
+```
 @inproceedings{Gantner2011MyMediaLite,
   author    = {Zeno Gantner and Steffen Rendle and Christoph Freudenthaler and Lars Schmidt-Thieme},
   title     = {{MyMediaLite}: A Free Recommender System Library},
   booktitle = {5th ACM International Conference on Recommender Systems (RecSys 2011)},
   year      = 2011,
   location  = {Chicago, USA}
 }
+```
+
+
+[Academic publications that use or reference MyMediaLite](https://scholar.google.de/scholar?q=mymedialite)
 
-Please inform us about if you have a publication that uses
-this software, so we can list it on MyMediaLite's website:
-Just send an e-mail to mymedialite@googlegroups.com
 
 Contributors
 ------------
+
 Thanks to the following people, who provided valuable feedback, code, or other
 kinds of assistance:
 Roberto Abalde, Nicholas Ampazis, Thorsten Angermann, Suhrid Balakrishnan,
@@ -74,21 +79,23 @@ Simon Dooms, Lucas Drumond, Michael Ekstrand, Christoph Freudenthaler,
 Zeno Gantner, Jagadeesh Gorla, Josif Grabocka, Mark Graus, Guibing Guo,
 Andreas Hoffmann, Tomas Horvath, Kenneth Hoste, Frantisek Hrdina, Jia Huang,
 Nicolas Hug, Dominik Imrich, Dietmar Jannach, Peng Jiang, KwangSeob Kim,
-Artus Krohn-Grimberghe, Tobias Lang, Christina Lichtenthäler, Damir Logar
+Artus Krohn-Grimberghe, Tobias Lang, Christina Lichtenthäler, Damir Logar,
 Marcelo Manzato, Brian McFee, Greg Najda, Chris Newell, Thai-Nghe Nguyen,
 Dimitris Paraschakis, Simon Renaud, Steffen Rendle, Marco Ribeiro,
 Roland Richter, Saurabh S., Sebastian Schelter, Lars Schmidt-Thieme, Yue Shi,
-Jordan Silva, Jessica Tölke, Tom Tung,Pieter-Jan Verbruggen, Julien Verplanken,
-Elvio Vicosa, João Vinagre, Oleksandr Vitvitskyi, Yongfeng Wang,
-Lina Weichbrodt, Cees Wesseling, Yong Zheng, GitHub users jkleint and NEBabylon.
+Jordan Silva, Piotr Sobotka, Jessica Tölke, Tom Tung, Pieter-Jan Verbruggen,
+Julien Verplanken, Elvio Vicosa, João Vinagre, Oleksandr Vitvitskyi,
+Yongfeng Wang, Lina Weichbrodt, Cees Wesseling, Yong Zheng, GitHub users
+jkleint and NEBabylon.
 
 This work was funded by the European Commission FP7 project MyMedia
-(Dynamic Personalization of Multimedia, http://www.mymediaproject.org/)
-under the grant agreement no. 215006.
+(Dynamic Personalization of Multimedia) under the grant agreement no. 215006.
+
 
 Copyright & Licensing
 ---------------------
-Copyright (C) 2010, 2011, 2012, 2013, 2015, 2016 The MyMediaLite contributors
+
+Copyright (C) 2010, 2011, 2012, 2013, 2015, 2016, 2019 The MyMediaLite contributors
 
 MyMediaLite is free software: you can redistribute it and/or modify
 it under the terms of the GNU General Public License as published by
@@ -104,14 +111,7 @@ You should have received a copy of the GNU General Public License
 along with MyMediaLite.  If not, see <http://www.gnu.org/licenses/>.
 
 ----
-This package contains Mono.Options, C5, the Math.NET Numerics library,
-a C# port of LIBSVM, and NUnit.
-
-The original packages can be found here:
-http://www.itu.dk/research/c5/
-http://numerics.mathdotnet.com
-http://www.csie.ntu.edu.tw/~cjlin/libsvm/
-http://www.matthewajohnson.org/software/svm.html
-http://www.nunit.org
+This package contains Mono.Options, [C5](http://www.itu.dk/research/c5/), [Math.NET Numerics](http://numerics.mathdotnet.com),
+and [NUnit](http://www.nunit.org).
 
 See doc/ComponentLicenses for more information about their licensing terms.

doc/Changes

@@ -1,9 +1,19 @@
-3.13	2017-??-??:
+3.13	2019-??-??:
         recommenders:
          - IIterativeModel: get rid of ComputeObjective() -- not used
          - remove LatentFeatureLogLinearModel -- implementation was never complete
+         - #423 #479: speed up external recommenders (Petr Sobotka)
+        command-line tools:
+         - allow modifying the floating point format with --prediction-line
 	hyperparameter search:
          - do not explore negative values of hyperparameters
+        evaluation:
+         - ensure culture-invariant output
+        documentation:
+         - convert README to markdown format
+        misc
+         - support Mono 5.x, drop support for older versions of Mono
+         - dataset download: update URLs, also download ML-20M
 
 3.12	2016-12-30
 	recommenders:

doc/Installation

@@ -1,5 +1,5 @@
 Requirements for running MyMediaLite:
- - Mono 2.8.x or another recent .NET runtime (Mono 2.10.x recommended)
+ - Mono 5.x or another recent .NET runtime (Mono 5.16 recommended)
  - Perl and File::Slurp for the download and data processing scripts
    (not necessary for running/building MyMediaLite)
 
@@ -19,18 +19,8 @@ Requirements for building MyMediaLite:
    - Graphviz
 
 Tested configurations:
- - Ubuntu 10.04
-   - with manually installed Mono 2.10.5 and MonoDevelop 2.6/2.8.1
-   - Doxygen 1.6.3
- - Ubuntu 11.10
-   - with Mono 2.10.5
- - Ubuntu 14.04
-   - with Mono 3.2.8 and MonoDevelop 4.0.12
- - Ubuntu 16.04 LTS
-   - with Mono 4.2.1 and MonoDevelop 5.10
- - Debian Wheezy
-   - with Mono 2.10.1 and MonoDevelop 2.8.2
- - Windows with Visual Studio 2010
+ - Ubuntu 18.04 LTS
+   - with Mono 5.16
 
 Let us know on which configuration of .NET or Mono you are using MyMediaLite,
 or whether you experience difficulties.

examples/csharp/Makefile

@@ -9,4 +9,4 @@ run:
 	#mono --debug foldin_incremental.exe
 
 clean:
-	rm -f *.exe *.mdb
+	rm -f *.exe *.pdb

scripts/download_imdb.sh

@@ -2,6 +2,8 @@
 
 # to be run from the root directory of MyMediaLite
 
+set -e
+
 echo "The IMDB data is for NON-COMMERCIAL use only."
 echo "Refer to http://www.imdb.com/Copyright for the details of the usage terms."
 echo "Hit enter to continue, Ctrl-C to abort."
@@ -10,14 +12,13 @@ read DUMMY
 echo "This may take a while ..."
 echo
 
-mkdir data/imdb
+mkdir -p data/imdb
+cd data/imdb
 
 # download IMDB data
-wget --output-document=data/imdb/german-aka-titles.list.gz ftp://ftp.fu-berlin.de/pub/misc/movies/database/german-aka-titles.list.gz
-
-cd data/imdb
+wget https://datasets.imdbws.com/title.akas.tsv.gz
 
 # unzip data
-gunzip german-aka-titles.list.gz
+gunzip title.akas.tsv.gz
 
 cd ../..

scripts/download_movielens.sh

@@ -2,23 +2,28 @@
 
 # to be run from the root directory of MyMediaLite
 
+set -e
+
 echo "The MovieLens datasets are for NON-COMMERCIAL use only."
 echo "Refer to the README files for the details of the usage license."
 
 cd data
 
-# download the other 2 MovieLens datasets
-wget http://grouplens.org/system/files/ml-1m.zip
-wget http://files.grouplens.org/papers/ml-10m.zip
+
+# download the other 3 MovieLens datasets
+wget http://files.grouplens.org/datasets/movielens/ml-1m.zip
+wget http://files.grouplens.org/datasets/movielens/ml-10m.zip
+wget http://files.grouplens.org/datasets/movielens/ml-20m.zip
 
 # unzip data
-unzip ml-1m.zip
-unzip ml-10m.zip
+unzip -a ml-1m.zip
+unzip -a ml-10m.zip
+unzip -a ml-20m.zip
 
 mv ml-10M100K ml-10m
 
 # remove downloaded archives
-rm ml-1m.zip ml-10m.zip
+rm ml-1m.zip ml-10m.zip ml-20m.zip
 
 
 # create attribute files for MovieLens 1M
@@ -35,4 +40,9 @@ rm ml-1m.zip ml-10m.zip
 ../scripts/import_dataset.pl --separator=:: ml-10m/ratings.dat > ml-10m/ratings.txt
 ../scripts/ml1m_genres.pl ml-10m/movies.dat > ml-10m/item-attributes-genres.txt
 
+# create tab-separated file for MovieLens 20M
+../scripts/import_dataset.pl --separator=, --ignore-lines=1 ml-20m/ratings.csv > ml-20m/ratings.txt
+../scripts/ml1m_genres.pl --ignore-lines=1 ml-20m/movies.csv > ml-20m/item-attributes-genres.txt
+
+
 cd ..

scripts/import_dataset.pl

@@ -62,7 +62,7 @@
 
 print STDERR "Start with $user_count and $item_count known users and items\n";
 
-# skip lines
+# skip lines at beginning of file
 for (my $i = 0; $i < $ignore_lines; $i++) { <>; }
 
 LINE:

scripts/ml1m_genres.pl

@@ -1,6 +1,6 @@
 #!/usr/bin/perl
 
-# (c) 2009, 2010, 2011 by Zeno Gantner
+# (c) 2009, 2010, 2011, 2019 by Zeno Gantner
 #
 # This file is part of MyMediaLite.
 #
@@ -28,7 +28,8 @@
 
 GetOptions(
 	   'help'           => \(my $help          = 0),
-	   'sparse-output!' => \(my $sparse_output = 1),
+           'sparse-output!' => \(my $sparse_output = 1),
+           'ignore-lines=i' => \(my $ignore_lines  = 0),
 	  ) or usage(-1); ## TODO implement usage
 
 my @genres = (
@@ -45,16 +46,19 @@
 $genre_id{Children} = $genre_id{"Children's"};
 $genre_id{IMAX}     = $counter++;
 
+# skip lines at beginning of file
+for (my $i = 0; $i < $ignore_lines; $i++) { <>; }
+
 while (<>) {
     my $line = $_;
     chomp $line;
 
-    my @fields = split '::', $line;
-    if (scalar @fields != 3) {
-	croak "Could not parse line: '$line'\n";
+    unless ($line =~ /^(\d+).+?([-A-Za-z'| ()]*)$/) {
+	die "Could not parse line: '$line'\n";
     }
 
-    my ($movie_id, $movie_title, $movie_genres) = @fields;
+    my $movie_id = $1;
+    my $movie_genres = $2;
 
     next if $movie_genres eq '(no genres listed)';
 
@@ -67,7 +71,7 @@
 		print "$movie_id\t$genre_id\n";
 	    }
 	    else {
-		print STDERR "Unknown genre in line '$line'\n";
+		print STDERR "Unknown genre in line '$line': '$movie_genres'\n";
 	    }
 	}
     }

src/Makefile

@@ -2,10 +2,10 @@ all: ../lib/mymedialite
 	xbuild /p:Configuration=Debug MyMediaLite-nogui.sln
 	cp -r Programs/ItemRecommendation/bin/Debug/*.dll ../lib/mymedialite
 	cp -r Programs/ItemRecommendation/bin/Debug/*.exe ../lib/mymedialite
-	cp -r Programs/ItemRecommendation/bin/Debug/*.mdb ../lib/mymedialite
-	cp -r Programs/RatingPrediction/bin/Debug/*.mdb ../lib/mymedialite
+	cp -r Programs/ItemRecommendation/bin/Debug/*.pdb ../lib/mymedialite
+	cp -r Programs/RatingPrediction/bin/Debug/*.pdb ../lib/mymedialite
 	cp -r Programs/RatingPrediction/bin/Debug/*.exe ../lib/mymedialite
-	cp -r Programs/RatingBasedRanking/bin/Debug/*.mdb ../lib/mymedialite
+	cp -r Programs/RatingBasedRanking/bin/Debug/*.pdb ../lib/mymedialite
 	cp -r Programs/RatingBasedRanking/bin/Debug/*.exe ../lib/mymedialite
 	cp -r Programs/OnlineEvaluator/bin/Debug/*.exe ../lib/mymedialite
 	cp -r Programs/OnlineEvaluator/bin/Debug/*.mdb ../lib/mymedialite