User contributions for StackMoreLayers

Search for contributionsshowhide

⧼contribs-top⧽

Username, IP address or CIDR range:

Namespace:

Invert selection

Associated namespace

Tag filter:

Invert selection

Only show edits that are latest revisions

Only show edits that are page creations

Hide minor edits

⧼contribs-date⧽

From date:

To date:

Show only edits pending review

Hide probably good edits

16:4216:42, 4 April 2024 diff hist −144 Simpson's paradox →Psychology: Snip a bit of text insinuating that the sure-thing principle (which is a theorem!) is a misconception. That's misleading. Maybe this was trying to say that people try to apply causal reasoning to non-causal contingency tables, baffling themselves and psychologists alike with their fervent attachment to causality, or some such, but if so, it failed.

10:1510:15, 8 October 2022 diff hist +393 Maze-solving algorithm →Trémaux's algorithm: Clarify the algorithm explanation. Use "entrance" as a name for junction-passage pairs instead of "path". Make it clear when marks are placed ("as you pass") and when entrance choices are made ("when you are at a junction") so that it is clear which marks are in place when you're counting them. Mention that the illustration does not match the text in that regard.

04:4304:43, 31 July 2022 diff hist +172 Broyden–Fletcher–Goldfarb–Shanno algorithm Mention that the Wolfe conditions entail the curvature condition when we first mention that something special might need to be done to satisfy it.
03:4803:48, 31 July 2022 diff hist +392 Symmetric rank-one The equations are busy enough that it's easy to miss why positive-definiteness is not preserved. Add a quick explanation for it (basically, the denominator can be negative).
03:2803:28, 31 July 2022 diff hist +11 Symmetric rank-one SR1 is typically derived by assuming a symmetric rank-1 update, which leads to a unique solution. "Closest" (in the sense of some norm) is typically used when deriving DFP or BFGS. How does "closest" work for SR1?

05:0405:04, 22 January 2022 diff hist −501 Fisher consistency →Relationship to asymptotic consistency and unbiasedness: These are not Fisher consistent. The concept does not even apply since they lack a representation as a functional of the ECDF.

12:0412:04, 18 January 2022 diff hist +79 DKW Add hatnote link to the DKW inequality in statistics.

23:2423:24, 16 January 2022 diff hist +338 Wisdom of the crowd Mention jury theorems and social choice theory.

15:4815:48, 13 January 2022 diff hist +6 Euclidean distance →Squared Euclidean distance: Oops, dropped a factor of two. Tag: Reverted
15:1915:19, 13 January 2022 diff hist −13 Cosine similarity →Definition: The vectors are assumed to be nonzero, no need for this caveat after all.
15:1715:17, 13 January 2022 diff hist +683 Euclidean distance →Squared Euclidean distance: Relation to dot product and cosine distance. Tag: Reverted
13:5713:57, 13 January 2022 diff hist +77 Cosine similarity →Definition: Euclidean distance (suitably used) preserves order, too, as explained in the L2 section.
13:5213:52, 13 January 2022 diff hist −3 m Cosine similarity TOC does not like math tags in titles, resulting in a mess.
13:5013:50, 13 January 2022 diff hist +282 Cosine similarity →Definition: Reduce the failure of triangle inequality to squared Euclidean distance, which is more common and possibly more intuitive than cosine distance.
13:3113:31, 13 January 2022 diff hist +99 Cosine similarity →L_2-normalised Euclidean distance: Add a note on why order is preserved.
13:0613:06, 13 January 2022 diff hist +176 Cosine similarity →Properties: Give the explicit formula for cosine distance in terms of Euclidean distance along with the normalization conditions in one easy to notice displaymath box.

10:4710:47, 12 January 2022 diff hist +923 Legendre symbol →Computational example: At least mention a polynomial time algorithm since we're talking about efficient computation.

18:2218:22, 30 December 2021 diff hist +1,658 Combination Mention the multinomial coefficient generalization
11:2411:24, 30 December 2021 diff hist +169 Combination →Probability: sampling a random combination: Mention the combinatorial number system for sampling, too (it's already mentioned for enumeration, but it's a rather different application)

20:2420:24, 3 November 2021 diff hist +83 Cosine similarity →Properties: Link the polarization identity to its main article

19:3919:39, 11 April 2021 diff hist +12 Singular value decomposition →Compact SVD: Update to match thin SVD section, so this does not assume that n <= m.
18:2818:28, 11 April 2021 diff hist +17 m Singular value decomposition →Thin SVD: Oops, accidentally dropped <sub>k</sub>. Also give shape for V_k^* instead of V_k to match the compact/truncated SVD sections.
18:2218:22, 11 April 2021 diff hist +771 Singular value decomposition →Thin SVD: Mention the symmetric case with fewer rows than columns. Rewrite using min(m, n) and max(m, n) so that there's no need to go through each symmetric case separately. Add citation.
04:3704:37, 11 April 2021 diff hist +203 Singular value decomposition It is not true that all solutions are real if M is real. "Real SVD" produces real outputs because it is assumed that solutions exiting the real field are invalid, but you may also compute a "complex SVD" of a real matrix, and then a complex solution is valid.
02:0102:01, 11 April 2021 diff hist +1,027 Polar decomposition Add a definition for rectangular matrices.
00:5900:59, 11 April 2021 diff hist +51 Orthogonal matrix →Rectangular matrices: Link to the [[semi-orthogonal matrix
00:5000:50, 11 April 2021 diff hist +279 Semi-orthogonal matrix Add another book reference (1011 citations on Google Scholar), and also a paper available on the Web (227 citations). This term is clearly in use.

22:0022:00, 10 April 2021 diff hist +101 N User:StackMoreLayers ←Created page with '{{Image frame|content=800px|align=center|caption=Design sketch.}}' current