Weighted Density sampling is a method to discard events based on their local density. Please read our introduction to sampling if you have not done so already to learn about your options for calculating local density.

 

Weighted density sampling is implemented as originally described in Kollios, G et. al "Efficient Biased Sampling for Approximate Clustering and Outlier Detection in Large Data Sets", IEEE Transactions on Knowledge and Data Engineering, 15(5):1170-1187, 2003.

 

There are several options for Weighted density sampling:

 

Sample Size

Indicates how many events to sample.

Alpha

Alpha is used to calculate the local density. Please refer to the Local Density section of the Sampling Options chapter for more details.

Weight

The Weight parameter controls the probability with which an events is sampled based on the local density. Positive weights mean that events from higher density regions will be preferentially sampled. Negative weights mean that events from low density regions will be preferentially sampled. A weight of 0 means random sampling will be performed, with no bias for density. In general values from -3 to 3 are appropriate. In order to mimic Target Density sampling, positive values between 0 - 1 are generally appropriate.

Local Density Method

Local Density is further described in the Local Density section of the Sampling Options chapter.

Add Local Density as Output

Check this box to create a new parameter that reports the local density value for each event.

Local Density Output Name

If Add Local Density as Output is checked, the user can specify a name for the new parameter.

www.denovosoftware.com