Kernel Density Estimation and Non-parametric Bayes Classifier

Estimating a probability density function is one of the fundamental tasks in statistical inference. A probability density function of a continuous random variable describes the relative likelihood of the variable at each point in the observation space. In general, a point with a low probability density represents a rare event, which in some cases corresponds to an outlier. The probability density of a point can also be used to decide if it belongs to certain class of points.

Our implementation offers several important features:

• Fast computation of kde for big query and reference dataset with trees.
• Approximate computation of KDE based on a time budget (progressive mode)
• Approximate computation of KDE with deterministic error bounds
• Approximate computation of KDE with probabilistic error bounds
• Automatic bandwidth selection
• Gaussian and Epanechnikov kernel support
• Non-parametric Bayes classifier based on KDE

Related articles

• Kernel Density Estimation
  • Choosing the Right Bandwidth
  • Making KDE fast
  • 1305 Implementation of KDE
  • Planned Extensions of Kernel Density Estimation Implemented in 1305
• Running KDE

Downloading

You can download it from here

Usage examples

• Running KDE
• A Simple Example
• Bichromatic KDE
• Gaussian, Epanechnikov and relative error
• KDE with kd-trees, metric-trees (ball-trees)
• KDE for sparse data
• KDE in progressinve mode (on a budget)
• Kernel Discriminant Analysis (KDA), using KDE for classification

Suggestions and bug reports

We would like to know your opinion about our product. Please send us bug reports and suggestions at support@analytics1305.com

If you didn't find an answer to your question, check the forum http://www.analytics1305.com/forum