In high-D cases (somewhere between D=100-250), the empirical (MLE) sample covariance estimator becomes very unstable. I need to implement one of these more robust estimators to avoid this problem and ensure reasonable bounding ellipsoids. This will add some optional dependencies that I should flag for users who want to apply dynesty to very high-D distributions.
In high-D cases (somewhere between D=100-250), the empirical (MLE) sample covariance estimator becomes very unstable. I need to implement one of these more robust estimators to avoid this problem and ensure reasonable bounding ellipsoids. This will add some optional dependencies that I should flag for users who want to apply
dynestyto very high-D distributions.