Apply service components

[jemus42]

This is part of the effort to unify the setups between lecture_* repositories.

Main goal is to establish a shared collection of resources in slds-lmu/lecture_service (see also slds-lmu/teaching_devops_issues/issues/6)

Summary

This PR

• Applies the style folder as collected in the service repo, closely based on I2ML, with notable changes:
  • Addition of "render slides with 16:9 format and speaker margin" as default behavior
  • "Render 4:3 slides without margin" as secondary option (achieved via Makefile)
  • \date{} is no longer required to be set to empty in slides, as date field is removed from slide footer
  • Same goes for \author, which is commented out in most slides anyway
• Updates /slides/tex.mk which is the Makefile used in /slides/<topic>/ folders
  • Now has make help target for some documentation (still WIP)
  • As most-normargin and most targets to render slides without or with speaker margin.
• Rearranges the lecture_sl.tex file to render all slides in one PDF, see below
• Add /scripts/update-service.sh which downloads the service directory from the service repo and copies it ontop of the repo for update-without-deleting service files (workflows, Makefiles, /style/preamble.tex, ...)
• Adds minor infrastructure bits and bobs from the service repo (unified .gitignore, README files, workflows)

To do in this repo

• [x] Ensure all slides compile in main
• [x] Check changes to style etc don't mess up slides
• [x] Check tex.mk / slides/*/Makefilecan still be used as intended
• Update slides in ./slides-pdf/ if necessary to squash spurious dissimilarities

In other repos

Nothing so far

Collateral damage

The lecture_sl.tex file should compile a "full lecture"-style PDF, which in I2ML is done within slides/all and in the interest of consistency I moved it from slides-pdf (where it is unexpected, I'd argue) to slides/all and tried adjusting all paths as needed, but currently it doesn't compile correctly as the other slides/<topic>/chapter-order.tex files assume a different path for the slides-pdf/<slide-name>.pdf files.

These paths starting with ../slides-pdf should apparently start with ../../slides-pdf, which is done here as well.

I realize a little late that this is essentially https://github.com/slds-lmu/teaching_devops_issues/issues/7 which needs fixing anyway.

Why did I not notice that this is just #128.

[github-actions[bot]]

Lecture Slide Status Overview

2024-04-10 15:17:11 CEST

• lecture_sl
  • advriskmin
  • boosting
  • cod
  • feature-selection
  • information-theory
  • linear-svm
  • lookup-slides
  • mathrefresher
  • multiclass
  • nonlinear-svm
  • regularization

This is a reduced version of the full status check rendered for each lecture repository.
It does not link to rendered PDF files and only produces markdown tables, which allows it to be appended to a pull request. It serves as a rough check to ensure that the PR does not cause major breakage.

Branch	Last Commit	By	Message
main	2024-04-10 12:25:02	Lukas Burk	Makefile tweaks

Status of the lecture_service repo (time in UTC)

Lecture	Branch	Last Commit	By	Message
lecture_sl	?	2024-04-10 13:09:06	GitHub	Merge 9cc23f08a0ffc4e7882fb5c5e5ef97dc8c9b9ce4 into 1ba1834d2192441861a83fa7cf0f51b99ca7f544

Latest commits per lecture (times in UTC)

lecture_sl

11 topics with a total of 83 slides

compile_check	n
✅	83

compile_check	compare_check	n
✅		1
✅	✅	35
✅	❌	4
✅	❓	43

advriskmin

Slide	Compiles?	Comparison	Note
slides-advriskmin-bias-variance-decomposition-deepdive	✅	❓	Dissimilar pages: 2, 4, 5
slides-advriskmin-bias-variance-decomposition	✅	❓	Dissimilar pages: 2, 11, 13
slides-advriskmin-classification-01	✅	✅
slides-advriskmin-classification-bernoulli	✅	✅
slides-advriskmin-classification-brier	✅	✅
slides-advriskmin-classification-deepdive	✅	✅
slides-advriskmin-classification-furtherlosses	✅	✅
slides-advriskmin-logreg-deepdive	✅	❓	Dissimilar pages: 6
slides-advriskmin-losses-properties	✅	❓	Dissimilar pages: 7, 9
slides-advriskmin-max-likelihood-l2	✅	✅
slides-advriskmin-max-likelihood-other	✅	✅
slides-advriskmin-pseudo-residuals	✅	✅
slides-advriskmin-regression-further-losses	✅	✅
slides-advriskmin-regression-l1-deepdive	✅	✅
slides-advriskmin-regression-l2-l1	✅	✅
slides-advriskmin-risk-minimizer	✅	❌	Differing page count: 13 vs 11

boosting

Slide	Compiles?	Comparison	Note
slides-boosting-cwb-advanced	✅	✅
slides-boosting-cwb-basics	✅	✅
slides-boosting-cwb-basics2	✅	❓	Dissimilar pages: 6
slides-boosting-cwb-glm	✅	✅
slides-boosting-gbm-classification	✅	❓	Dissimilar pages: 4
slides-boosting-gbm-regularization	✅	✅
slides-boosting-gbm-with-trees-1	✅	❓	Dissimilar pages: 4, 5, 11, 12, 13
slides-boosting-gbm-with-trees-2	✅	❓	Dissimilar pages: 3
slides-boosting-gradient-boosting-concept	✅	❓	Dissimilar pages: 4, 11, 15, 16
slides-boosting-intro-adaboost	✅	❓	Dissimilar pages: 12
slides-boosting-lgm-ctbm	✅	❓	Dissimilar pages: 3
slides-boosting-regression-illustrations	✅	✅
slides-boosting-xgboost-deepdive	✅	✅
slides-boosting-xgboost	✅	❓	Dissimilar pages: 5

cod

Slide	Compiles?	Comparison	Note
slides-cod-examples	✅	❓	Dissimilar pages: 3, 4, 6, 8, 10
slides-cod	✅	❓	Dissimilar pages: 4, 5, 17, 18, 19

feature-selection

Slide	Compiles?	Comparison	Note
slides-fs-filters1	✅	❓	Dissimilar pages: 2, 9
slides-fs-filters2	✅	❓	Dissimilar pages: 2, 3, 5
slides-fs-introduction	✅	✅
slides-fs-motivating-examples	✅	✅
slides-fs-wrapper	✅	❓	Dissimilar pages: 1, 3

information-theory

Slide	Compiles?	Comparison	Note
slides-info-cross-entropy-kld	✅	❓	Dissimilar pages: 8
slides-info-diffent	✅	✅
slides-info-entropy	✅	✅
slides-info-entropy2	✅	❓	Dissimilar pages: 7
slides-info-kl-ment	✅	❓	Dissimilar pages: 3, 4, 5, 7
slides-info-kl-ml	✅	❓	Dissimilar pages: 7
slides-info-kl	✅	❓	Dissimilar pages: 11
slides-info-mi-deepdive	✅	✅
slides-info-ml	✅	✅
slides-info-mutual-info	✅	✅
slides-info-mutual-info2	✅	✅
slides-info-sourcecoding	✅	❓	Dissimilar pages: 3
slides-info-sourcecoding2	✅	✅

linear-svm

Slide	Compiles?	Comparison	Note
slides-linsvm-erm	✅	❌	Differing page count: 7 vs 6
slides-linsvm-hard-margin-dual	✅	❓	Dissimilar pages: 6
slides-linsvm-hard-margin	✅	✅
slides-linsvm-optimization	✅	❓	Dissimilar pages: 7
slides-linsvm-soft-margin	✅	✅

lookup-slides

Slide	Compiles?	Comparison	Note
riskmin_optconst	✅		No reference PDF

mathrefresher

Slide	Compiles?	Comparison	Note
slides-probability-theory	✅	❓	Dissimilar pages: 14, 16, 17, 21, 23

multiclass

Slide	Compiles?	Comparison	Note
slides-mc-binary-reduction	✅	❓	Dissimilar pages: 3, 6, 8, 9, 11
slides-mc-codebooks	✅	❓	Dissimilar pages: 3, 4, 5, 9, 11
slides-mc-losses	✅	❓	Dissimilar pages: 3, 7
slides-mc-softmax-regression	✅	❓	Dissimilar pages: 4, 5, 6, 7, 11

nonlinear-svm

Slide	Compiles?	Comparison	Note
slides-nonlinsvm-featuregen	✅	❌	Differing page count: 9 vs 8
slides-nonlinsvm-kernel-poly	✅	❓	Dissimilar pages: 2, 3
slides-nonlinsvm-kernel-rbf	✅	❓	Dissimilar pages: 3, 12
slides-nonlinsvm-kernel-trick	✅	❓	Dissimilar pages: 2, 4, 9, 12, 13
slides-nonlinsvm-modelsel	✅	✅
slides-nonlinsvm-rkhs-repr	✅	❓	Dissimilar pages: 3, 5, 6, 9
slides-nonlinsvm-uniapprox	✅	❌	Differing page count: 10 vs 8

regularization

Slide	Compiles?	Comparison	Note
slides-regu-bagging-deepdive	✅	❓	Dissimilar pages: 6
slides-regu-bayes	✅	✅
slides-regu-bias-variance	✅	❓	Dissimilar pages: 2, 3, 5
slides-regu-early-stopping	✅	✅
slides-regu-enetlogreg	✅	✅
slides-regu-geom-l1	✅	✅
slides-regu-geom-l2-wdecay	✅	❓	Dissimilar pages: 11, 13
slides-regu-intro	✅	❓	Dissimilar pages: 5
slides-regu-l1	✅	✅
slides-regu-l1vsl2	✅	❓	Dissimilar pages: 7
slides-regu-l2	✅	❓	Dissimilar pages: 11
slides-regu-lasso-deepdive	✅	❓	Dissimilar pages: 3, 4, 5, 6
slides-regu-nonlin	✅	❓	Dissimilar pages: 4, 5, 6, 7, 8
slides-regu-others	✅	✅
slides-regu-ridge-deepdive	✅	❓	Dissimilar pages: 3, 4, 5, 6

[jemus42]

As far as I can tell this isn't breaking anything and the PDF diffs seem fine, or at least not too different than the ones from the main branch anyway (I'm checking that locally).

@chriskolb are you in charge of reviewing/merging?