Skip to main content

Table 8 SER, error per class and average class error on the test partition for the BLSTM1PoolBLSTM2 RNN classifier (Avg pooling, N=10) trained using mixup augmentation with hyperparameter α

From: Multiclass audio segmentation based on recurrent neural networks for broadcast domain data

MixupSERClass error (%)Avg
  muspsmsn 
α=0.114.8415.2127.9923.0529.3423.90
α=0.214.8014.6428.2022.0129.0123.56
α=0.314.8116.0326.3223.8928.2223.62
No mixup15.4715.5529.1624.3430.5124.89