Timeline for PCA vs linear Autoencoder: features independence
Current License: CC BY-SA 4.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 28, 2020 at 19:03 | history | edited | AJKOER | CC BY-SA 4.0 |
added 22 characters in body
|
| May 28, 2020 at 18:59 | comment | added | Sycorax♦ | Keep in mind that the edit history of the post will always be available, so you don't have to worry about discarding what you originally wrote. If someone wants to see earlier versions, they can find it easily using the edit history. Just a tip. | |
| May 28, 2020 at 18:57 | comment | added | AJKOER | Sycorax: Agree, clumsy, but trying to edit for correctness and maintain original words. I also added some of the applications' success stories as background differentiation from PCA. | |
| May 28, 2020 at 18:51 | history | edited | AJKOER | CC BY-SA 4.0 |
added 995 characters in body
|
| May 28, 2020 at 18:45 | comment | added | Sycorax♦ | Your third edit states "...there is necessarily no 'independence', at least, in a Linear Algebra sense," which appears to suggest that the resulting feature vectors are not linearly independent as a matter of necessity. I think you meant to write "there is not necessarily independence in a linear algebra sense," which means that the result may or may not be linear independent vectors. | |
| May 28, 2020 at 18:41 | comment | added | AJKOER | Sycorax: I will accept your point, and have clearly placed Edit and [END EDIT] in my answer. I doubt, however, in my opinion, as to whether it significantly distracts from the key point of my analysis. Supporting my opinion, I have further added an extract from Wikipedia as provided background . | |
| May 28, 2020 at 18:32 | history | edited | AJKOER | CC BY-SA 4.0 |
added 87 characters in body
|
| May 28, 2020 at 15:18 | comment | added | Sycorax♦ | I think you're misreading the quotation. "[T]he new features we end up do not have to be independent" means that the new features might be independent or might not be independent, but instead you write "the new features data set is neither orthogonal or even linearly independent." It's true that they're not orthogonal, but vectors can be linearly independent and also be non-orthogonal. An example is shown in my answer. | |
| May 28, 2020 at 14:11 | history | answered | AJKOER | CC BY-SA 4.0 |