generate pdf version of tutorial

.travis.yml

@@ -22,6 +22,7 @@ addons:
       - g++-4.9
       - gfortran-4.9
       - libgmp-dev
+      - pandoc
 
 before_install:
   - mkdir $HOME/bin
@@ -38,7 +39,7 @@ install:
   - Rscript r_packs_install.R
 
 script:
-  - ./build || ./build || ./build || ./build
+  - ./build
 
 after_success:
     - git checkout gh-pages
@@ -48,6 +49,7 @@ after_success:
     - echo "https://$GH_TOKEN:@github.com" >> .git/credentials
     - git config push.default matching
     - git add devel
+    - git add mlr-tutorial.pdf
     - git commit -a -m "update auto-generated tutorial pages [ci skip]"
     - git push
 

README.md

@@ -34,10 +34,10 @@ Install dependencies:
 * Add the new section to the pages configuration in `mkdocs.yml`.
 
 ### Include images
-If you want to include an additional image in file `pic.png`:
-* Put this file in subfolder `images/`.
-* Add a symlink in directory `custom_theme/img/`: `pic.png -> ../../images/pic.png`.
-* When including the image in the R markdown link to `img/pic.png`:
+If you want to include an additional image in file `pic.png`:  
+* Put this file in subfolder `img/`.
+* Add a symlink in directory `custom_theme/img/`: `pic.png -> ../../img/pic.png`.
+* When including the image in the R markdown link to `img/pic.png`:  
   `![alt text](img/pic.png "Image Title")`
 
 ### Commit your changes

build

@@ -77,14 +77,14 @@ setSeed = function(x, algo = "crc32") {
 }
 
 # worker function
-knitIt = function(f) {
+knitIt = function(f, dev) {
   messagef("Knitting file '%s' ...", f)
   lines = readLines(file.path(src, f))
   for (macro in macros)
     lines = str_replace_all(lines, macro$pattern, macro$replacement)
 
   cache.path = paste0(file.path("cache", f), .Platform$file.sep)
-  opts_chunk$set(cache = TRUE, cache.path = cache.path, dev = "svg", error = FALSE, comment = "#>", collapse = TRUE)
+  opts_chunk$set(cache = TRUE, cache.path = cache.path, fig.path = paste0(file.path("figure", str_replace(f, "\\.Rmd", "")), .Platform$file.sep), dev = dev, error = FALSE, comment = "#>", collapse = TRUE)
   setSeed(f)
   knit(
     text = lines,
@@ -116,28 +116,28 @@ suppressPackageStartupMessages({
 
 # turn warnings to errors, we don't want to miss them
 options(warn = 2L)
-# embed images in HTML
-opts_knit$set(upload.fun = image_uri)
 # number of spaces to indent the code is 2 (for chunks with tidy=TRUE)
-opts_chunk$set(tidy.opts=list(indent=2, width.cutoff=80))
+opts_chunk$set(tidy.opts = list(indent = 2, width.cutoff = 80))
 
 # create output directory
 if (!isDirectory("docs"))
   dir.create("docs")
 
 # do lapply for now. with mclapply it seems unclear how to stop best on error. this was 100% broken before
-dummy = lapply(list.files(src, pattern = "\\.Rmd$"), knitIt)
-
-# build docs with mkdocs
-ok = system3("mkdocs", "build")
+dummy = lapply(list.files(src, pattern = "\\.Rmd$"), knitIt, "pdf")
+message("Building PDF version...")
+ok = system3("./build-pdf.py")
 if (ok$exit.code == 0L) {
-  message("Build successfull!")
+  # embed images in HTML
+  opts_knit$set(upload.fun = image_uri)
+  message("Building HTML version...")
+  dummy = lapply(list.files(src, pattern = "\\.Rmd$"), knitIt, "svg")
+  ok = system3("mkdocs", "build")
+  message("Build successful!")
   message("Now zip everything!")
   setwd("devel/html/")
   suppressAll(zip("../mlr_tutorial.zip", files = "."))
-  setwd("../")
-  # clean up
-  unlink("figure", recursive = TRUE)
+  setwd("../../")
   exit(0L)
 }
 

build-pdf.py

@@ -0,0 +1,64 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+import os
+import re
+import sys
+
+os.chdir("docs")
+with open(os.path.join("..", "mlr-tutorial.md"), "w") as fd:
+  fd.write("""
+---
+title: mlr Tutorial
+title-meta: mlr Tutorial
+author:
+  - Julia Schiffner
+  - Bernd Bischl
+  - Michel Lang
+  - Jakob Richter
+  - Zachary M. Jones
+  - Philipp Probst
+  - Florian Pfisterer
+  - Mason Gallo
+  - Dominik Kirchhoff
+  - Tobias Kühn
+  - Lars Kotthoff
+---
+""")
+  with open(os.path.join("..", "mkdocs.yml"), "r") as fd2:
+    line = fd2.readline()
+    while line:
+      if('Appendix' in line):
+        break
+      elif('.md' in line):
+        m = re.search("[^']+\.md", line)
+        fname = m.group(0)
+        with open(fname, "r") as fd3:
+          for lin in fd3:
+            if fname != "index.md":
+              lin = re.sub(r'^(#+ [A-Za-z0-9])', '#\g<1>', lin)
+            fd.write(lin)
+          fd.write("\n")
+        if('index.md' in line):
+          fd.write("\n\n# Basics\n\n")
+        if('visualization.md' in line):
+          fd.write("\n\n# Advanced\n\n")
+        if('hyperpar_tuning_effects.md' in line):
+          fd.write("\n\n# Extend\n\n")
+      line = fd2.readline()
+os.chdir("..")
+
+def link_fixer(match):
+  file = match.group(1)
+  with open(os.path.join("docs", file), 'r') as fd:
+    return "(" + re.sub(' ', '-', re.sub(' ', '', fd.readline().rstrip(), 1)).lower() + ")"
+
+with open('mlr-tutorial.md', 'r+') as fd:
+  data = fd.read()
+  out = re.sub(r'\(([a-zA-Z0-9_]+\.md)\)', link_fixer, data)
+  fd.seek(0)
+  fd.write(out)
+  fd.truncate()
+
+retval = os.system("pandoc --number-sections --latex-engine=xelatex --variable colorlinks=\"true\" --listings -H latex-setup.tex --toc -f markdown+grid_tables+table_captions-implicit_figures -o mlr-tutorial.pdf mlr-tutorial.md")
+if retval != 0:
+  sys.exit(1)

custom_theme/img/mlrLogo_white_88x40.png

@@ -1 +1 @@
-../../images/mlrLogo_white_88x40.png
+../../img/mlrLogo_white_88x40.png

custom_theme/img/nested_resampling.png

@@ -1 +1 @@
-../../images/nested_resampling.png
+../../img/nested_resampling.png

custom_theme/img/resampling.png

@@ -1 +1 @@
-../../images/resampling.png
+../../img/resampling.png

devel/html/advanced_tune/index.html

@@ -380,12 +380,12 @@ <h2 id="iterated-f-racing-for-mixed-spaces-and-dependencies">Iterated F-Racing f
 #&gt; 5 -10.079158    polydot       NA      5           0.06   1  NA
 #&gt; 6 -10.643475 vanilladot       NA     NA           0.36   1  NA
 #&gt;   error.message exec.time
-#&gt; 1          &lt;NA&gt;     0.039
-#&gt; 2          &lt;NA&gt;     0.025
-#&gt; 3          &lt;NA&gt;     0.024
-#&gt; 4          &lt;NA&gt;     0.041
-#&gt; 5          &lt;NA&gt;     0.040
-#&gt; 6          &lt;NA&gt;     0.038
+#&gt; 1          &lt;NA&gt;     0.049
+#&gt; 2          &lt;NA&gt;     0.024
+#&gt; 3          &lt;NA&gt;     0.025
+#&gt; 4          &lt;NA&gt;     0.025
+#&gt; 5          &lt;NA&gt;     0.024
+#&gt; 6          &lt;NA&gt;     0.024
 </code></pre>
 
 <p>See how we made the kernel parameters like <code>sigma</code> and <code>degree</code> dependent on the <code>kernel</code>
@@ -435,12 +435,12 @@ <h2 id="tuning-across-whole-model-spaces-with-modelmultiplexer">Tuning across wh
 #&gt; 5         classif.ksvm          -1.884101                         NA
 #&gt; 6         classif.ksvm           4.388728                         NA
 #&gt;   mmce.test.mean dob eol error.message exec.time
-#&gt; 1      0.6466667   1  NA          &lt;NA&gt;     0.042
-#&gt; 2      0.1400000   1  NA          &lt;NA&gt;     0.042
-#&gt; 3      0.0400000   1  NA          &lt;NA&gt;     0.059
-#&gt; 4      0.0400000   1  NA          &lt;NA&gt;     0.034
-#&gt; 5      0.0400000   1  NA          &lt;NA&gt;     0.042
-#&gt; 6      0.3333333   1  NA          &lt;NA&gt;     0.042
+#&gt; 1      0.6466667   1  NA          &lt;NA&gt;     0.047
+#&gt; 2      0.1400000   1  NA          &lt;NA&gt;     0.044
+#&gt; 3      0.0400000   1  NA          &lt;NA&gt;     0.063
+#&gt; 4      0.0400000   1  NA          &lt;NA&gt;     0.036
+#&gt; 5      0.0400000   1  NA          &lt;NA&gt;     0.043
+#&gt; 6      0.3333333   1  NA          &lt;NA&gt;     0.044
 </code></pre>
 
 <h2 id="multi-criteria-evaluation-and-optimization">Multi-criteria evaluation and optimization</h2>
@@ -473,12 +473,12 @@ <h2 id="multi-criteria-evaluation-and-optimization">Multi-criteria evaluation an
 #&gt; 5 5.203364e+00 2.781734e-03     0.1515152     0.1621622   5  NA
 #&gt; 6 5.638243e-04 7.956946e+02     1.0000000     0.0000000   6  NA
 #&gt;   error.message exec.time
-#&gt; 1          &lt;NA&gt;     0.067
-#&gt; 2          &lt;NA&gt;     0.069
-#&gt; 3          &lt;NA&gt;     0.069
-#&gt; 4          &lt;NA&gt;     0.071
-#&gt; 5          &lt;NA&gt;     0.067
-#&gt; 6          &lt;NA&gt;     0.064
+#&gt; 1          &lt;NA&gt;     0.042
+#&gt; 2          &lt;NA&gt;     0.043
+#&gt; 3          &lt;NA&gt;     0.043
+#&gt; 4          &lt;NA&gt;     0.044
+#&gt; 5          &lt;NA&gt;     0.042
+#&gt; 6          &lt;NA&gt;     0.043
 </code></pre>
 
 <p>The results can be visualized with function <a href="http://www.rdocumentation.org/packages/mlr/functions/plotTuneMultiCritResult.html">plotTuneMultiCritResult</a>.

devel/html/configureMlr/index.html

@@ -511,7 +511,7 @@ <h2 id="example-handling-errors-in-a-learning-method">Example: Handling errors i
 <pre><code class="r">## This call gives an error caused by the low number of observations in class &quot;virginica&quot;
 train(&quot;classif.qda&quot;, task = iris.task, subset = 1:104)
 #&gt; Error in qda.default(x, grouping, ...): some group is too small for 'qda'
-#&gt; Timing stopped at: 0.004 0 0.004
+#&gt; Timing stopped at: 0.003 0 0.003
 
 ## Get a warning instead of an error
 configureMlr(on.learner.error = &quot;warn&quot;)

devel/html/create_filter/index.html

@@ -386,7 +386,7 @@ <h2 id="filter-objects">Filter objects</h2>
 #&gt;     y = FSelector::rank.correlation(getTaskFormula(task), data = getTaskData(task))
 #&gt;     setNames(y[[&quot;attr_importance&quot;]], getTaskFeatureNames(task))
 #&gt; }
-#&gt; &lt;bytecode: 0x9c89308&gt;
+#&gt; &lt;bytecode: 0x8000410&gt;
 #&gt; &lt;environment: namespace:mlr&gt;
 </code></pre>
 

devel/html/feature_selection/index.html

@@ -499,19 +499,19 @@ <h3 id="tuning-the-size-of-the-feature-subset">Tuning the size of the feature su
 #&gt; With control class: TuneControlGrid
 #&gt; Imputation value: Inf
 #&gt; [Tune-x] 1: fw.perc=0.2
-#&gt; [Tune-y] 1: mse.test.mean=40.6; time: 0.0 min; memory: 131Mb use, 645Mb max
+#&gt; [Tune-y] 1: mse.test.mean=40.6; time: 0.0 min; memory: 131Mb use, 623Mb max
 #&gt; [Tune-x] 2: fw.perc=0.25
-#&gt; [Tune-y] 2: mse.test.mean=40.6; time: 0.0 min; memory: 131Mb use, 645Mb max
+#&gt; [Tune-y] 2: mse.test.mean=40.6; time: 0.0 min; memory: 131Mb use, 623Mb max
 #&gt; [Tune-x] 3: fw.perc=0.3
-#&gt; [Tune-y] 3: mse.test.mean=37.1; time: 0.0 min; memory: 131Mb use, 645Mb max
+#&gt; [Tune-y] 3: mse.test.mean=37.1; time: 0.0 min; memory: 131Mb use, 623Mb max
 #&gt; [Tune-x] 4: fw.perc=0.35
-#&gt; [Tune-y] 4: mse.test.mean=35.8; time: 0.0 min; memory: 131Mb use, 645Mb max
+#&gt; [Tune-y] 4: mse.test.mean=35.8; time: 0.0 min; memory: 131Mb use, 623Mb max
 #&gt; [Tune-x] 5: fw.perc=0.4
-#&gt; [Tune-y] 5: mse.test.mean=35.8; time: 0.0 min; memory: 131Mb use, 645Mb max
+#&gt; [Tune-y] 5: mse.test.mean=35.8; time: 0.0 min; memory: 131Mb use, 623Mb max
 #&gt; [Tune-x] 6: fw.perc=0.45
-#&gt; [Tune-y] 6: mse.test.mean=27.4; time: 0.0 min; memory: 131Mb use, 645Mb max
+#&gt; [Tune-y] 6: mse.test.mean=27.4; time: 0.0 min; memory: 131Mb use, 623Mb max
 #&gt; [Tune-x] 7: fw.perc=0.5
-#&gt; [Tune-y] 7: mse.test.mean=27.4; time: 0.0 min; memory: 131Mb use, 645Mb max
+#&gt; [Tune-y] 7: mse.test.mean=27.4; time: 0.0 min; memory: 131Mb use, 623Mb max
 #&gt; [Tune] Result: fw.perc=0.5 : mse.test.mean=27.4
 res
 #&gt; Tune result:
@@ -522,13 +522,13 @@ <h3 id="tuning-the-size-of-the-feature-subset">Tuning the size of the feature su
 <p>The performance of all percentage values visited during tuning is:</p>
 <pre><code class="r">as.data.frame(res$opt.path)
 #&gt;   fw.perc mse.test.mean dob eol error.message exec.time
-#&gt; 1     0.2      40.59578   1  NA          &lt;NA&gt;     0.199
-#&gt; 2    0.25      40.59578   2  NA          &lt;NA&gt;     0.194
-#&gt; 3     0.3      37.05592   3  NA          &lt;NA&gt;     0.193
-#&gt; 4    0.35      35.83712   4  NA          &lt;NA&gt;     0.194
-#&gt; 5     0.4      35.83712   5  NA          &lt;NA&gt;     0.195
-#&gt; 6    0.45      27.39955   6  NA          &lt;NA&gt;     0.190
-#&gt; 7     0.5      27.39955   7  NA          &lt;NA&gt;     0.192
+#&gt; 1     0.2      40.59578   1  NA          &lt;NA&gt;     0.240
+#&gt; 2    0.25      40.59578   2  NA          &lt;NA&gt;     0.267
+#&gt; 3     0.3      37.05592   3  NA          &lt;NA&gt;     0.271
+#&gt; 4    0.35      35.83712   4  NA          &lt;NA&gt;     0.314
+#&gt; 5     0.4      35.83712   5  NA          &lt;NA&gt;     0.278
+#&gt; 6    0.45      27.39955   6  NA          &lt;NA&gt;     0.229
+#&gt; 7     0.5      27.39955   7  NA          &lt;NA&gt;     0.237
 </code></pre>
 
 <p>The optimal percentage and the corresponding performance can be accessed as follows:</p>
@@ -571,12 +571,12 @@ <h3 id="tuning-the-size-of-the-feature-subset">Tuning the size of the feature su
 #&gt; Points on front: 13
 head(as.data.frame(res$opt.path))
 #&gt;   fw.threshold fpr.test.mean fnr.test.mean dob eol error.message exec.time
-#&gt; 1    0.4892321     0.3092818     0.2639033   1  NA          &lt;NA&gt;     1.914
-#&gt; 2    0.2481696     0.2045499     0.2319697   2  NA          &lt;NA&gt;     1.937
-#&gt; 3    0.7691875     0.5128000     0.3459740   3  NA          &lt;NA&gt;     1.824
-#&gt; 4    0.1470133     0.2045499     0.2319697   4  NA          &lt;NA&gt;     1.922
-#&gt; 5    0.5958241     0.5028216     0.5239538   5  NA          &lt;NA&gt;     1.816
-#&gt; 6    0.6892421     0.6323959     0.4480808   6  NA          &lt;NA&gt;     1.826
+#&gt; 1    0.4892321     0.3092818     0.2639033   1  NA          &lt;NA&gt;     2.991
+#&gt; 2    0.2481696     0.2045499     0.2319697   2  NA          &lt;NA&gt;     2.778
+#&gt; 3    0.7691875     0.5128000     0.3459740   3  NA          &lt;NA&gt;     2.035
+#&gt; 4    0.1470133     0.2045499     0.2319697   4  NA          &lt;NA&gt;     2.023
+#&gt; 5    0.5958241     0.5028216     0.5239538   5  NA          &lt;NA&gt;     1.835
+#&gt; 6    0.6892421     0.6323959     0.4480808   6  NA          &lt;NA&gt;     2.022
 </code></pre>
 
 <p>The results can be visualized with function <a href="http://www.rdocumentation.org/packages/mlr/functions/plotTuneMultiCritResult.html">plotTuneMultiCritResult</a>.

devel/html/hyperpar_tuning_effects/index.html

@@ -414,14 +414,14 @@ <h2 id="generating-hyperparameter-tuning-data">Generating hyperparameter tuning
 #&gt; Measures: acc.test.mean,mmce.test.mean
 #&gt; Optimizer: TuneControlRandom
 #&gt; Nested CV Used: FALSE
-#&gt; Snapshot of data:
+#&gt; Snapshot of $data:
 #&gt;            C acc.test.mean mmce.test.mean iteration exec.time
-#&gt; 1  0.3770897     0.7695312      0.2304688         1     0.074
+#&gt; 1  0.3770897     0.7695312      0.2304688         1     0.065
 #&gt; 2  3.4829323     0.7526042      0.2473958         2     0.065
-#&gt; 3  2.2050176     0.7630208      0.2369792         3     0.066
-#&gt; 4 24.9285221     0.7070312      0.2929688         4     0.073
-#&gt; 5  0.2092395     0.7539062      0.2460938         5     0.067
-#&gt; 6  0.1495099     0.7395833      0.2604167         6     0.068
+#&gt; 3  2.2050176     0.7630208      0.2369792         3     0.064
+#&gt; 4 24.9285221     0.7070312      0.2929688         4     0.074
+#&gt; 5  0.2092395     0.7539062      0.2460938         5     0.066
+#&gt; 6  0.1495099     0.7395833      0.2604167         6     0.066
 </code></pre>
 
 <p>As a reminder from the <a href="../resample/index.html">resampling</a> tutorial, if we wanted to generate data on
@@ -440,7 +440,7 @@ <h2 id="generating-hyperparameter-tuning-data">Generating hyperparameter tuning
 #&gt; Measures: acc.test.mean,acc.train.mean,mmce.test.mean,mmce.train.mean
 #&gt; Optimizer: TuneControlRandom
 #&gt; Nested CV Used: FALSE
-#&gt; Snapshot of data:
+#&gt; Snapshot of $data:
 #&gt;            C acc.test.mean acc.train.mean mmce.test.mean mmce.train.mean
 #&gt; 1 0.03518875     0.6510417      0.6510417      0.3489583       0.3489583
 #&gt; 2 0.17104229     0.7356771      0.7721354      0.2643229       0.2278646
@@ -449,12 +449,12 @@ <h2 id="generating-hyperparameter-tuning-data">Generating hyperparameter tuning
 #&gt; 5 1.28168692     0.7500000      0.8476562      0.2500000       0.1523438
 #&gt; 6 7.36607693     0.7239583      0.8932292      0.2760417       0.1067708
 #&gt;   iteration exec.time
-#&gt; 1         1     0.093
-#&gt; 2         2     0.094
-#&gt; 3         3     0.091
-#&gt; 4         4     0.092
-#&gt; 5         5     0.089
-#&gt; 6         6     0.091
+#&gt; 1         1     0.092
+#&gt; 2         2     0.089
+#&gt; 3         3     0.088
+#&gt; 4         4     0.089
+#&gt; 5         5     0.090
+#&gt; 6         6     0.088
 </code></pre>
 
 <p>In the example below, we perform grid search on the <code>C</code> parameter for SVM on the Pima Indians
@@ -476,14 +476,14 @@ <h2 id="generating-hyperparameter-tuning-data">Generating hyperparameter tuning
 #&gt; Measures: acc.test.mean,mmce.test.mean
 #&gt; Optimizer: TuneControlGrid
 #&gt; Nested CV Used: TRUE
-#&gt; Snapshot of data:
+#&gt; Snapshot of $data:
 #&gt;            C acc.test.mean mmce.test.mean iteration exec.time
-#&gt; 1 -5.0000000     0.6640625      0.3359375         1     0.049
-#&gt; 2 -3.8888889     0.6640625      0.3359375         2     0.049
-#&gt; 3 -2.7777778     0.6822917      0.3177083         3     0.049
-#&gt; 4 -1.6666667     0.7473958      0.2526042         4     0.049
-#&gt; 5 -0.5555556     0.7708333      0.2291667         5     0.050
-#&gt; 6  0.5555556     0.7682292      0.2317708         6     0.050
+#&gt; 1 -5.0000000     0.6640625      0.3359375         1     0.087
+#&gt; 2 -3.8888889     0.6640625      0.3359375         2     0.086
+#&gt; 3 -2.7777778     0.6822917      0.3177083         3     0.085
+#&gt; 4 -1.6666667     0.7473958      0.2526042         4     0.087
+#&gt; 5 -0.5555556     0.7708333      0.2291667         5     0.084
+#&gt; 6  0.5555556     0.7682292      0.2317708         6     0.085
 #&gt;   nested_cv_run
 #&gt; 1             1
 #&gt; 2             1

devel/html/index.html

@@ -437,5 +437,5 @@ <h4 class="modal-title" id="exampleModalLabel">Search</h4>
 
 <!--
 MkDocs version : 0.15.3
-Build Date UTC : 2016-09-15 10:44:20.978412
+Build Date UTC : 2016-08-30 12:32:08.083858
 -->