Metrics

AbstractMetrics

Every metrics struct is a subtype of AbstractMetrics

Accuracy

Represents the classic accuracy metric

BinaryCrossentropy

Represents the binary accuracy metric

Field

• threshold: The threshold used to decide if the output is 0 or 1. Every predictions > threshold is set to 1

CategoricalAccuracy

Represents the categorical accuracy metric

compute_accuracy(metrics::BinaryAccuracy, predictions::Tensor, target::Union{Tensor,AbstractArray,Float64,Int64})
compute_accuracy(metrics::CategoricalAccuracy, predictions::Tensor, target::Union{Tensor,AbstractArray,Float64,Int64})
compute_accuracy(metrics::Accuracy, predictions::Tensor, target::Union{Tensor,AbstractArray,Float64,Int64})

Compute the accuracy according to the metrics given.