Talk:Variational autoencoder

On the recent additions

@~2025-27630-11 the Simple English Wikipedia's first and foremost goal is to be an English encyclopedia that is easily understandable by many people. This is accomplished by avoiding complex language and writing using Simple English. Your edit has several glaring issues with it.

First, your language is not encyclopedic at all. Sentences like which both look at the same side, pointing at the same direction! could do without the exclamation mark. Sentences like But we said that we have two hourglass halves that pour their sand into one. How do we build such a thing? only serve to pander the reader and are of no encyclopedic value. Use a disinterested, neutral tone that does not try to captivate the reader needlessly. Don't ask questions to the reader. The Simple English Wikipedia is an encyclopedia. It is not a Medium blog.

Second, your language is far from simple. What is "mathematical pressure" in they use mathematical pressure through the "cost function"? Why is it only explained much later in your changes? And it makes sense because otherwise the random number would be at the beginning of our shenanigans and it would confuse the mathematical pressure. is a very complex sentence. Try to break down complex sentences to simple sentences: sentences that have a subject, predicate and an object (where needed). There are cases where this is insufficient but for most of your explanations you should aim to write in this way.

Third, you should add sources to your edits. You added four paragraphs and none of them have any sources.

If you have trouble grasping what writing Simple English is like, then take a look WP:HOW for the structure, and WP:BE1500 for the vocabulary. It is your responsibility to make your edits simple. You should not leave it to others to simply your edits. I am reverting your edit again until you have cleaned up your edit. — *Fehufangą ♮ ✉ Talk page 01:58, 6 October 2025 (UTC)

Variational autoencoders are extremely complex and you have not done what I have asked you, which was to check the primary article to understand the magnitude of the complexity that we are talking about. You can judge my writing all you want, but I find it irresponsible to delete things that cannot be easily reproduced. Even the first edit admits the issues, and you can choose to bring up policy if you want. But deleting content instead of helping make it better does not make you a good editor.

Finally, I reiterate that there are very few people who can simplify the topic as much as I did. You can choose to make it better, or to throw it out. Choose the wikipedia you like. If you do choose the former, the only reference that you need to add is Kingma and Welling's Autoencoding Variational Bayes, which is the source I use. ~2025-27630-11 (talk) 15:31, 6 October 2025 (UTC)

@~2025-27630-11 I know how complex these are because I have studied them in university. I know how complex it is trying to explain deep learning in a simple concept.

but I find it irresponsible to delete things that cannot be easily reproduced except we do remove content that we deem to complex in this wiki: https://quarry.wmcloud.org/query/97812 should give a list of some of these changes. Your revision is no exception. You should not expect the work of simplification to be done by others. You added the bulk of the content, expanding the article by over two thousand bytes. Strive for simplicity, not florid and complex language. — *Fehufangą ♮ ✉ Talk page 07:31, 7 October 2025 (UTC)

I have greatly simplified a complex topic while maintaining its nuance. If you wish to ignore my request to edit together and decide to throw out my work then so be it. ~2025-27630-11 (talk) 11:23, 7 October 2025 (UTC)

I will eventually circle back to the topic. If someone wants to collaborate with a person who has authored peer-reviewed research in prestigious international journals, feel free to give a go at my deleted edit. I will check it out when I have time to discuss misconceptions such as variational autoencoders being deep learning models (just look at probabilistic pca for a counterexample) ~2025-27630-11 (talk) 17:24, 12 October 2025 (UTC)