Facebook Transcoder : Unsupervised Translation of Programming Languages | Paper Explained

#NLProc #Facebook #AI
In this video I will explain about Facebook Transcoder : Unsupervised Translation of Programming Languages. Transcoder is a deep learning based unsupervised translation model which can translate programs from programming languages like C++ to Java or Python and vice-versa. This model was developed by researchers at Facebook. Manual code translation is very expensive and laborious process. This paper describes a fully unsupervised machine translation approach to solve this problem.

If you are new here and like such content please subscribe to the channel here:
https://www.youtube.com/c/RitheshSreenivasan?sub_confirmation=1

If you like to support me financially, It is totally optional and voluntary.
Buy me a coffee here: https://www.buymeacoffee.com/rithesh

Chapters
00:00 Facebook Transcoder : Unsupervised Translation of Programming Languages
00:45 Need for unsupervised translation of programming languages
02:14 Main Contributions
03:16 Transcoder Model
03:49 Cross-lingual Masked Language Model pretraining
05:12 Denoising auto-encoding
08:08 Back-translation
10:06 Training details
11:11 Training data
12:23 Preprocessing
13:30 Evaluation
17:10 Evaluation Metrics
19:10 Results
21:57 Cross-lingual token embedding space
24:28 Summary
Unsupervised Translation of Programming Languages:
https://arxiv.org/pdf/2006.03511.pdf

To learn more about Transformer models:
http://jalammar.github.io/illustrated-transformer/

To learn more about seq-seq models with attention:
http://jalammar.github.io/visualizing-neural-machine-translation-mechanics-of-seq2seq-models-with-attention/

Видео Facebook Transcoder : Unsupervised Translation of Programming Languages | Paper Explained канала Rithesh Sreenivasan