Talk:Matrix multiplication algorithm: Difference between revisions

Content deleted Content added
actually, this is high importance
DDavo (talk | contribs)
 
(12 intermediate revisions by 7 users not shown)
Line 1:
{{WikiProject Computing}}banner shell|class=C|
{{WikiProject ComputerComputing science|importance=highLow}}
{{WikiProject Computer science |importance=Mid }}
}}
{{merged from|Cache-oblivious matrix multiplication}}
{{merged-from|Coppersmith–Winograd algorithm|May 20, 2021}}
 
== Numerical stability ==
 
This article is still missing information about numerical stability. In fact, this important topic is only briefly mentioned when Strassen's algo is introduced. [http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=E8428C66852DF918AAFCDACFC9738D37?doi=10.1.1.148.9947&rep=rep1&type=pdf This article] (already cited in the text) might be a starting point. [[User:Qwertyus|Q<small>VVERTYVS</small>]] <small>([[User talk:Qwertyus|hm?]])</small> 18:42, 17 August 2015 (UTC)
 
== row-major order? ==
 
....the above algorithm is sub-optimal for A and B stored in row-major order...
 
Why? <math>sum \leftarrow sum + A_{ik} \times B_{kj}</math>
 
Loop for k in A runs in row by columns, in B runs in column by rows... [[User:Jumpow|Jumpow]] ([[User talk:Jumpow|talk]]) 13:39, 3 March 2021 (UTC)
 
== Merger proposal ==
{{Discussion top|result=The result of this discussion was '''merge'''.}}
 
I propose to merge [[Coppersmith–Winograd algorithm]] into [[Matrix multiplication algorithm]]. The [[Matrix_multiplication_algorithm#Sub-cubic_algorithms|sub-cubic algorithms]] section contains almost all the information in [[Coppersmith–Winograd algorithm]], and most of the articles that link to Coppersmith–Winograd are either referring to the fastest matrix multiplication algorithm (which is no long Coppersmith-Winograd) or simply any fast matrix multiplication algorithm. I've also seen cases where people don't realize that a matrix multiplication algorithm article exists, and find the article for Coppersmith-Winograd instead.
 
As another related point, it seems like [[Computational complexity of matrix multiplication]] could be made its own article, considering that both [[Matrix multiplication algorithm]] and [[Matrix multiplication]] contain one section about the usual type of computational complexity of matrix multiplication, with the rest of their articles discussing other things entirely (implementation details and mathematical definitions, respectively). I'm not as sure about this, though. [[User:Fawly|Fawly]] ([[User talk:Fawly|talk]]) 23:02, 30 April 2021 (UTC)
 
: I support the merge. [[User:Saung Tadashi|Saung Tadashi]] ([[User talk:Saung Tadashi|talk]]) 01:27, 1 May 2021 (UTC)
{{Discussion bottom}}
 
== MIT OpenCourseWare hyperlink stale: plug slides or video? ==
 
The hyperlink to <em>MIT OpenCourseWare 6.172</em> has gone stale:<br/>
plug [https://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-172-performance-engineering-of-software-systems-fall-2018/lecture-slides/MIT6_172F18_lec1.pdf slides] or video?
[[Special:Contributions/94.220.51.20|94.220.51.20]] ([[User talk:94.220.51.20|talk]]) 07:47, 13 November 2021 (UTC)
 
== Winograd's cubic algorithm ==
 
I just learned that Winograd's has another algorithm, from before Strassen's that does half the multiplications (It's O(0.5n³))
 
https://dl.acm.org/doi/10.1109/TC.1968.227420 [[User:DDavo|DDavo]] ([[User talk:DDavo|talk]]) 08:05, 23 May 2025 (UTC)