Talk:Projection pursuit regression: Difference between revisions
Adding class to statistics rating template |
→Model overview inconsistency: new section Tags: Mobile edit Mobile app edit Android app edit |
||
Line 2: | Line 2: | ||
Is there a typo here?: "It has also been shown that βj converges at an order of n^\frac{1}{2}, while βj converges at a slightly worse order." |
Is there a typo here?: "It has also been shown that βj converges at an order of n^\frac{1}{2}, while βj converges at a slightly worse order." |
||
== Model overview inconsistency == |
|||
if x_i is a row vector, the result of the matrix product in the argument of f_j cannot be a scalar. But f_j are scalar functions of scalar arguments. i believe the description of x_i is not correct. it should be something in the line of: x_i is a column vector formed from the i-th row of the design matrix. That is, it is the features vector. |
|||
Alternatively, if the idea is to stick to the rows of the design matrix, one should change the formula and write x_i beta_j without any transpose. To make all consistent one would state that the beta_j are column weight vectors [[User:Kakila|Kakila]] ([[User talk:Kakila|talk]]) 02:16, 4 November 2022 (UTC) |
Revision as of 02:16, 4 November 2022
![]() | Statistics Start‑class | |||||||||
|
Is there a typo here?: "It has also been shown that βj converges at an order of n^\frac{1}{2}, while βj converges at a slightly worse order."
Model overview inconsistency
if x_i is a row vector, the result of the matrix product in the argument of f_j cannot be a scalar. But f_j are scalar functions of scalar arguments. i believe the description of x_i is not correct. it should be something in the line of: x_i is a column vector formed from the i-th row of the design matrix. That is, it is the features vector. Alternatively, if the idea is to stick to the rows of the design matrix, one should change the formula and write x_i beta_j without any transpose. To make all consistent one would state that the beta_j are column weight vectors Kakila (talk) 02:16, 4 November 2022 (UTC)