PDF Projection Theorem and
Maximum Entropy PDF Estimation

WEBSITE UNDER CONSTRUCTION - Adding more content every day

Last updated Feb 18, 2017.

Link to PPT Home

The purpose of this website is to provide a central repository of information, software, models, and documentation concerning PDF estimation and classification using PDF projection and related methods. It is intended for those persons doing research in statistical signal processing, classification theory and methods.

Publications on the topic

Evaluating the RBM Without Integration Using PDF Projection (EUSIPCO2017 - submitted). Demonstrates how PDF projection can be used to analyze the RBM.

Uniform Manifold Sampling (UMS): Sampling the Maximum Entropy PDF, (IEEE Trans, SP 2017 Jan 2017). This follow-on article provides methods to sample from maximum entropy projected PDFs with applications in feature inversion.

Maximum Entropy PDF Projection: A Review (MaxEnt 2016 Conference, Ghent)

Maximum entropy feature fusion, (2016 19th International Conference on Information Fusion)

Maximum Entropy PDF Design Using Feature Density Constraints: Applications in Signal Processing , (IEEE Trans SP, vol 63, no. 11, 2015). This is the main article establishing the maximum entropy property of PDF projection.

Class-specific model mixtures for the classification of acoustic time series,(IEEE TAES, Vol 52, no. 4, 2016)

Class-specific model mixtures for the classification of time series,(EUSIPCO 2015)

Optimal detection and classification of diverse short-duration signals (IWCCSP 2014) . This article describes the chronological development of the theory starting with class-specific features.

The PDF Projection Theorem and the Class-Specific Method, (IEEE Trans, SP March, 2003). This is the main reference for the PDF projection theorem.

Multidimensional Probability Density Function Approximations for Detection, Classification, and Model Order Selection, (IEEE Trans, SP, Oct, 2001). This paper provides methods of evaluating the reference hypothesis PDF for many important features.

Feedback: Please email any technical or website questions or comments to p.m.baggenstoss@ieee.org