New, much tighter confidence intervals for the mean!
It is far tighter than so-called envelope methods and uses the same weak
set of assumptions. For the first time, it gives useful confidence intervals even when the sample size is small (less than 30).
You can download my code for RADICAL, my independent components analysis
algorithm, from the project page here.
Fast entropy estimation
The following is a very simple matlab routine, which, given a sample
from a one-dimensional random variable, estimates the entropy of
the distribution. It uses an entropy estimator closely related to
the one developed by Oldrich Vasicek in the 1970's. This entropy
estimator is at the heart of my fast ICA algorithm (RADICAL), and
has many applications.
Congealing is a process for jointly aligning a set of images, volumes,
or signals. A simple version of the code, demonstrating the technique
on binary images, is available from the project web page, which