SonataClusterCrossEntropyLoss<T>

Models & Types · AiDotNet.ComputerVision.Segmentation.PointCloud

Sonata’s cluster cross-entropy: -sum(target * log_softmax(student / tau)) per match, averaged within each scene and then across scenes (segment_coo(..., reduce="mean").mean()).

Implements: ISonataDistillationLoss<T>

Constructors

Constructor Summary
SonataClusterCrossEntropyLoss()

Methods

Method Summary
Tensor<T> Compute(Tensor<T> studentScores, Tensor<T> targets, double studentTemperature, int[] sampleOfMatch)