Workload distribution

Registration Board

If "Translation status" indicates WIP, the translator is still working on his translation. Even if a first version is available online. The translator may still have some things to adjust, such as the presentation of formulas or images, or does a proofreading of his work (it is common to make mistakes when writing in Markdown and you can't see what it does without have put a first version online). It is therefore useless to do PR to correct the work since it is still in progress. It is only once the status thinking of "OK" that a reviewer can start the revision.

Similarly, if the "Revision Status" says "WIP", the reviewer is still working. So there is no point in doing PR. A translation is considered finished once the "Revision Status" indicates "OK".

File name	Translator	Translation statut	Reviewer	Review statut
README.md	lbourdois	OK	lbourdois	OK
index.md	lbourdois	OK	lbourdois	OK
01.md	lbourdois	OK	lbourdois	OK
01-1.md	lbourdois	OK	lbourdois	OK
01-2.md	lbourdois	OK	lbourdois	OK
01-3.md	lbourdois	OK	lbourdois	OK
02.md	lbourdois	OK	lbourdois	OK
02-1.md	lbourdois	OK	lbourdois	OK
02-2.md	lbourdois	OK	lbourdois	OK
02-3.md	lbourdois	OK	lbourdois	OK
03.md	lbourdois	OK	lbourdois	OK
03-1.md	lbourdois	OK	lbourdois	OK
03-2.md	lbourdois	OK	lbourdois	OK
03-3.md	lbourdois	OK	lbourdois	OK
04.md	lbourdois	OK	lbourdois	OK
04-1.md	lbourdois	OK	lbourdois	OK
05.md	lbourdois	OK	lbourdois	OK
05-1.md	lbourdois	OK	lbourdois	OK
05-2.md	lbourdois	OK	lbourdois	OK
05-3.md	lbourdois	OK	lbourdois	OK
06.md	lbourdois	OK	lbourdois	OK
06-1.md	lbourdois	OK	lbourdois	OK
06-2.md	lbourdois	OK	lbourdois	OK
06-3.md	lbourdois	OK	lbourdois	OK
07.md	lbourdois	OK	lbourdois	OK
07-1.md	lbourdois	OK	lbourdois	OK
07-2.md	lbourdois	OK	lbourdois	OK
07-3.md	lbourdois	OK	lbourdois	OK
08.md	lbourdois	OK	lbourdois	OK
08-1.md	lbourdois	OK	lbourdois	OK
08-2.md	lbourdois	OK	lbourdois	OK
08-3.md	lbourdois	OK	lbourdois	OK
09.md	lbourdois	OK	lbourdois	OK
09-1.md	lbourdois	OK	lbourdois	OK
09-2.md	lbourdois	OK	lbourdois	OK
09-3.md	lbourdois	OK	lbourdois	OK
10.md	lbourdois	OK	lbourdois	OK
10-1.md	lbourdois	OK	lbourdois	OK
10-2.md	lbourdois	OK	lbourdois	OK
10-3.md	lbourdois	OK	lbourdois	OK
11.md	lbourdois	OK	lbourdois	OK
11-1.md	lbourdois	OK	lbourdois	OK
11-2.md	lbourdois	OK	lbourdois	OK
11-3.md	lbourdois	OK	lbourdois	OK
12.md	lbourdois	OK	lbourdois	OK
12-1.md	lbourdois	OK	lbourdois	OK
12-2.md	lbourdois	OK	lbourdois	OK
12-3.md	lbourdois	OK	lbourdois	OK
13.md	lbourdois	OK	lbourdois	OK
13-1.md	lbourdois	OK	lbourdois	OK
13-2.md	lbourdois	OK	lbourdois	OK
13-3.md	lbourdois	OK	lbourdois	OK
14.md	lbourdois	OK	lbourdois	OK
14-1.md	lbourdois	OK	lbourdois	OK
14-2.md	lbourdois	OK	lbourdois	OK
14-3.md	lbourdois	OK	lbourdois	OK
15.md	lbourdois	OK	lbourdois	OK
15-1.md	lbourdois	OK	lbourdois	OK
15-2.md	lbourdois	OK	lbourdois	OK
lecture01.sbv	lbourdois	OK
practicum01.sbv	lbourdois	OK
lecture02.sbv	lbourdois	OK
practicum02.sbv	lbourdois	OK
lecture03.sbv	lbourdois	OK
practicum03.sbv	lbourdois	OK
practicum04.sbv	lbourdois	OK
lecture05.sbv	lbourdois	OK
practicum05.sbv	lbourdois	OK
lecture06.sbv	lbourdois	OK
practicum06.sbv	lbourdois	OK
lecture07.sbv	lbourdois	OK
practicum07.sbv	lbourdois	OK
lecture08.sbv	lbourdois	OK
practicum08.sbv	lbourdois	OK
lecture09.sbv	lbourdois	OK
practicum09.sbv	lbourdois	OK
lecture10.sbv	lbourdois	OK
practicum10.sbv	lbourdois	OK
lecture11.sbv	lbourdois	OK
practicum11.sbv	lbourdois	OK
lecture12.sbv	lbourdois	OK
practicum12.sbv	lbourdois	OK
lecture13.sbv	lbourdois	OK
practicum13.sbv	lbourdois	OK
lecture14.sbv	lbourdois	OK
practicum14.sbv	lbourdois	OK
practicum15A.sbv	lbourdois	OK
practicum15B.sbv	lbourdois	OK

Indications

In the table above, you will find the list of files for which the French translation is currently missing. Or which need a review.

The .sbv files correspond to the translation of the subtitles of the course videos. As an indication, it takes at least 4 hours of work for a 1h30 video (this can be more for technical videos where mathematical formulas are developed for example). Keep this in mind when registering ! Attention! English subtitles have been artificially generated. They may therefore contain errors. A manual verification of the English texts is in progress (see the Caption Videos Wiki or the Slack for more information). Before translating, make sure that the .sbvfile in question has been manually checked!

The .md files correspond to the translation of the lecture notes taken by the students. It takes about 1 hour of work for one file (except for those presenting the content of the week's courses, which are much smaller). Attention! To preserve the homogeneity of the translation, it is requested that for a given file, the translator commits to do all the others of the week (so 4 files). The working time is then counted between 3 and 4 hours.

Terminology conventions

Here is the list of choices that have been retained to carry out the translation. It is the result of a discussion with Yann Le Cun.

For writing

The texts found on the website are course notes from students who attended Yann Le Cun's classes (and from the four guest speakers). For each course, several students were asked to put their own notes to be added to the site. Thus the notes in English came from about 130 different people, who did not work together to unify their text. Some people write in the past tense, while others write in the present tense. The abbreviations used are not necessarily always the same, people are not always designated in the same way, etc. The variety of authors also means that the way of writing varies from one file to another.

Thus the choices that have been made for French are as follows:

• use of the present tense in all the files (if the course is about a past event, as for example when Yann tells episodes about the history of artificial intelligence dating back to the 40's, the past tense is obviously used)
• Some students tend to write very long sentences of 5 or 6 lines. In order to keep the text from ending in Proust and to keep it understandable, the sentences have been simplified as much as possible (three sentences of 2 lines or instead of one of 6), and repetitions in the text have been removed.
• Concerning the technical vocabulary, it is detailed in the following sub-section.

For videos

Videos are not translated word for word. Indeed, in order to fluidify the translation and the comprehension, it was decided not to transcribe the "parasitic" filling and transition words ("you know", "sort of", "right", "so", etc.).

Terms

The translations that have been selected for the following list of words are the result of a discussion with Yann Le Cun :

Term	Translation	Reason
CNN	ConvNet	Yann is particularly keen to respect this translation
Convolutional sparse auto-encoder	Auto-encodeur épars convolutif	Yann's choice
Contrastive	Constrative au féminin, Contrastif au masculin	Yann's choice
Dropout	Dropout	Yann thought of the term "decimation" but it is already used in signal processing to mean "subsampling".
Energy-Based Models	Modèles à base d’énergie	Yann on the subject: "Not very satisfying, but I don't have better! "
Energy based factor graphs	Graphes de facteurs à base d'énergie	Yann's choice
Embedding	Enchâssement	Used by Yann in his book
Gated Graph ConvNets	Réseau à rupteur convolutif sur graphe	Yann's choice (it is an invention on his part)
Graph Attention Networks	Réseau à graphe d'attention	Yann's choice
Graph Convolution Networks	Réseau convolutif sur graphe (abréviation : GCN)	Yann's choice
Graph Domain	Domaine du graphe (comme le domaine de Fourier)	Yann's choice
Group Sparsity	Eparsité de groupe	Yann's choice
One hot	One hot	The notion of "canonical base vectors" could be used but seems a bit technical and the expression is rather long to translate barely 2 words. The choice was therefore made to let the English expression
Overfitting	Surentraînement	Yann's choice
Pooling	Pooling	Translations into French as "aggrégation", "agglomération", were not satisfactory
Residual Gated Graph Convolutional Network	Réseau à rupteur convolutif à résidu sur graphe	Yann's choice
Simple Energy-Based Factor Graphs with “Shallow” Factors	Graphes de facteur simples à base d'énergie avec facteurs peu profonds	Yann's choice
Sparse	Epars	Yann's choice
Sparsity	Eparsité	Yann's choice
Spectral graph convolutional neural networks	Réseau convolutif spectral sur graphe	Yann's choice
SplineGCNs / LapGCNs / CayleyNets / ChebNets / MoNets / GraphSage	SplineGCNs / LapGCNs / CayleyNets / ChebNets / MoNets / GraphSage	Easier to keep the English abbreviations according to Yann
Vanilla Spatial GCNs	Graphe ConvNet spatial ordinaire	Yann's choice
Yann LeCun	Yann Le Cun ou Yann	For the writing of the family name, the Breton form is adopted. Yann explains in his book why in Anglo-Saxon countries he is obliged to write his surname all attached. In the notes of the students in English, it is possible to find "Mr Yann LeCun", "Mr LeCun", "Doctor Yann LeCun", "Professor LeCun", etc. Yann wishes that in these texts, he is simply called Yann.