Talk:Segmentation-based object categorization: Difference between revisions

Revision as of 13:48, 4 May 2016

Latest revision as of 17:20, 3 April 2024

(4 intermediate revisions by 3 users not shown)

Line 1:

{{WikiProject Roboticsbanner shell|class=Stub|importance=low|attention=no|past-selected=no}}

{{WikiProject Robotics|importance=low|attention=no}}

}}

==Ncut==

In the section "The Ncut algorithm" the value "D" is introduced without explanation. Also, none of the "figures" actually occur in the article.

To follow this comment, the "ncut algorithm" section should refer/link to the [[degree_matrix|degree]] and [[adjacency_matrix|adjacency]] matrix of a graph (or even the [[Laplacian_matrix|Laplacian matrix]]) when defining D and W, as well as the [[Algebraic_connectivity|algebraic connectivity]] of the graph when referencing the second smallest eigenvalue. [[Special:Contributions/70.166.151.52|70.166.151.52]] ([[User talk:70.166.151.52|talk]]) 16:42, 15 February 2017 (UTC)

Are arbitrary values allowed for the weights (similarity measures), or are they restricted to some range (non-negative, [0,1], ...)? It seems to me that negative weights can potentially invert the sign of some of the fraction denominators causing semantic havoc. [[User:Mspreitz|Mspreitz]] ([[User talk:Mspreitz|talk]]) 02:40, 11 April 2011 (UTC)