Rank: 1681 based on 79 downloads (last 30 days) and 4 files submitted
photo

Marco B.

E-mail
Company/University
ECB / Frankfurt University

Personal Profile:

 

Watch this Author's files

 

Files Posted by Marco B. View all
Updated   File Tags Downloads
(last 30 days)
Comments Rating
12 Dec 2011 Screenshot Nearest positive semi-definite covariance matrix Find nearest positive semi-definite matrix to a symmetric matrix that is not positive semi-definite Author: Marco B. covariance matrix, statistics 14 2
  • 1.0
1.0 | 1 rating
27 May 2010 Screenshot Fan chart Fan chart Author: Marco B. fan chart, forecasting, density forecasts 35 0
26 May 2010 Screenshot EM Algorithm: i.i.d. Mixture Distribution Maximum likelihood estimation of parameters from 2-state i.i.d. Normal mixture Author: Marco B. em algorithm, iid mixture distribut... 10 0
24 May 2010 Forward Stepwise Regression Algorithm Forward stepwise model selection algorithm (FSRA) Author: Marco B. model selection, automatic model build..., forward stepwise regr... 20 1
  • 3.0
3.0 | 1 rating
Comments and Ratings on Marco B.'s Files View all
Updated File Comment by Comments Rating
16 Dec 2013 Nearest positive semi-definite covariance matrix Find nearest positive semi-definite matrix to a symmetric matrix that is not positive semi-definite Author: Marco B. Martin

Regarding Shuo Han's alternative solution. For robustness I had to modify it to the below. My theory is that the division with V sometimes introduces numerical errors that are large enough to result in a new negative eigenvalue.

[V,D] = eig(A);
A_psd = V * diag(max(diag(D),eps)) / V;

20 Sep 2012 Nearest positive semi-definite covariance matrix Find nearest positive semi-definite matrix to a symmetric matrix that is not positive semi-definite Author: Marco B. Shuo Han

This does not need fmincon() and can be done in two lines. Suppose the input matrix is A

[V,D] = eig(A);
A_psd = V * max(D,0) / V;

19 Jun 2011 Forward Stepwise Regression Algorithm Forward stepwise model selection algorithm (FSRA) Author: Marco B. Xiaolu

it will be better if it can output some results like rmse and regression value

Contact us