Talk:Unscented transform
Add topic | This article is rated Start-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | |||||||||||||
| ||||||||||||||
Correction
[edit]Sorry but in the example given with s1 s2 s3, the mean is (0,0) but the covariance matrix IS NOT Id.... and is not even diagonal...
Therealarty (talk) 10:50, 29 February 2012 (UTC)
If you are referring to the example with s1=(0,sqrt(2)), .... in section 3, then the mean is indeed zero and computing the sample covariance matrix yields the identity. — Preceding unsigned comment added by 79.231.110.144 (talk) 08:42, 14 May 2015 (UTC)
The linearised result in the example is not correct. Using this jacobian for cartesian to polar function: https://www.wolframalpha.com/input/?i=diff+{sqrt%28x^2%2By^2%29,atan%28y%2Fx%29}+wrt+x,y (needs to transposed as wolfram alpha doesn't show Jacobians the correct way round). and applying J * [12.3,7.6]' * J' = [1.84063 0.0448447; 0.0448447 0.011908] NOT [1.927,0.0443;0.0443;0.011] as stated in the example. — Preceding unsigned comment added by 62.189.28.130 (talk) 11:50, 1 March 2016 (UTC)
— First paragraph of section sigma points : Incorrect reference to the review of the variants of sets of sigma points by Menegaz et al [3], reference points incorrectly to a book by Krantz and al. — Preceding unsigned comment added by NonLynSys (talk • contribs) 16:22, 15 January 2018 (UTC)
— About reference : Zhang, W.; M. Liu; Z. Zhao (2009). "Accuracy Analysis of Unscented Transformation of Several Sampling Strategies". Proc. of the 10th Intl. Conf. on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing. ACIS. This paper does not show any numerical illustrations of its conclusion. It is also very short and concludes too quicly on its results. For these reason I would recommend removing this reference. NonLynSys (talk) 17:21, 15 January 2018 (UTC)
— There is a counter-example (f(x) = x^T x) to the assertion ″Any set of sigma points that encodes the mean and covariance correctly calculates the projected mean and covariance correctly to the second order.″ given in the paper "Some Relations Between Extended and Unscented Kalman Filters" by Gustaffson & Hendeby. — Preceding unsigned comment added by NonLynSys (talk • contribs) 13:07, 19 January 2018 (UTC)