evaluate_softmax: Evaluate matrices of true targets and predictions from layer...

In GenomeNet/deepG: Deep Learning for Genome Sequence Data

Evaluate matrices of true targets and predictions from layer with softmax activation.

Description

Compute confusion matrix, accuracy, categorical crossentropy and (optionally) AUC or AUPRC, given predictions and true targets. AUC and AUPRC only possible for 2 targets.

Usage

evaluate_softmax(y, y_conf, auc = FALSE, auprc = FALSE, label_names = NULL)

Arguments

y	Matrix of true target.
y_conf	Matrix of predictions.
auc	Whether to include AUC metric. Only possible for 2 targets.
auprc	Whether to include AUPRC metric. Only possible for 2 targets.
label_names	Names of corresponding labels. Length must be equal to number of columns of y.

Value

A list of evaluation results.

Examples

y <- matrix(c(1, 0, 0, 0, 1, 1), ncol = 2)
y_conf <- matrix(c(0.3, 0.5, 0.1, 0.7, 0.5, 0.9), ncol = 2)
evaluate_softmax(y, y_conf, auc = TRUE, auprc = TRUE, label_names = c("A", "B")) 

GenomeNet/deepG documentation built on Dec. 24, 2024, 12:11 p.m.