Unofficial implementation of 2021 Neurocomputing paper "VAE-based Deep SVDD for anomaly detection" TensorFlow 2.0 version
Dataset: MNIST
Normal class: 8 (For example)
Final Results:
R: 6.721636056900024
TP: 4279.0
TN: 937.0
FP: 37.0
FN: 4747.0
precision: 0.9914272244803701
recall: 0.474074889496179
f_measure: 0.6413892930514092
accuracy: 0.5215999947840001
auc_roc: 0.8635820952566418
Cannot reproduce similar results as paper's illustration
References: