Feature Vector Cluster Merging

Cluster Merging is a property of Feature Vector Clustering and Reidentifier that merges two or more clusters into one when they are more similar to each other than a configurable threshold. The current similarity threshold is defined by the initial threshold and the threshold_discount_rate parameters and decreases as the number of input samples in the potentially mergeable clusters grows. But never goes under the min_threshold parameter. For details about configurable parameters and calculating the current threshold of the Cluster Merging feature, see Configuring Feature Vector Clustering. The result of merging is a MERGE_EVENT shown in the FVClusterUpdateRecord of the concerning stream's Kafka topic. A new cluster is created with it's own cluster id and representative feature vector and the merged ones are deleted. For an example of a MERGE_EVENT type FVClusterUpdateRecord output, see Configuring Feature Vector Clustering.