Talk:ReRites

"Confusing" tag - separate from machine learning explanation or the same issue?
Hi! Thanks for engaging with this entry, User:JPxG. You added the "Multiple issues" tag - can you explain the "Confusing" tag please? Does it refer to the explanation of the machine learning, or is it a separate issue? Lijil (talk) 09:29, 12 October 2022 (UTC)

Explanation of the machine learning
Hi User:JPxG - you added a tag stating "This article needs attention from an expert in Machine Learning. The specific problem is: Explanation of the code makes no sense; TensorFlow, AWS and PyTorch are not neural networks. (October 2022)" I'm not sure what the problem is here, though. The text doesn't say that TensorFlow, AWS or PyTorch are neural networks, it quotes a statement from the author of the work explaining that it uses "neural network code (..) adapted from three corporate github-hosted machine-learning libraries: Tensorflow (Google), PyTorch (Facebook), and AWSD (SalesForce)". As I understand it the author trained a neural network using machine-learning libraries: Tensorflow (Google), PyTorch (Facebook), and AWSD (SalesForce) - there is no claim that the libraries are neural networks. Did you mean something different? Lijil (talk) 09:35, 12 October 2022 (UTC)
 * Sure. What I mean by this is that I don't think the description makes a lot of sense as a technical explanation of what he's using (PyTorch, for example, is a framework that doesn't itself specify architecture or implementation). The readme of his repository seems to mention a few models, although it's not quite clear which is being used for what tasks. It looks like the primary one is a LSTM/QRNN (the code allows either to be trained, and he doesn't really say which he used, so one hopes he's written about it in more detail somewhere else). The SalesForce model he implemented is described in this paper, which would probably be a good source for more technical details, but it is fairly late for me so I am not quite up to the task of diving in. jp×g 11:29, 12 October 2022 (UTC)