The following papers contain a typo for the formula for the sum of
squared differences in the derivation for the MLEs for the Asymmetric
Gaussian distribution. Although the papers contained a typo, the code
posted on the web (cited in those papers) contained the correct
formula and all experiments and analyses remain correct.
The correct formula contains a "2" and is:
- "Using Asymmetric Distributions to Improve Classifier Probabilities: A
Comparison of New and Standard Parametric Methods", Bennett
CMU-CS-02-126, Computer Science Department, School of Computer
Science, Carnegie Mellon University, April 2002. Section 3.2 and Line 22 of Appendix C.
- "Using Asymmetric Distributions to Improve Text Classifier Probability
Estimates", Bennett, SIGIR, 2003. Section 3.3.2.
- "Building Reliable Metaclassifiers for Text Learning",
Bennett, CMU-CS-06-121, Ph.D. Thesis, Computer Science Department, School of
Computer Science, Carnegie Mellon University, May 2006. Line 3.20 of page 37 in Chapter 3.
Paul Bennett, April 30, 2007.