Skip to main content

Table 5 First dataset’s results for different architectures and sequence length: training loss, validation accuracy, testing accuracy, and testing AUROC

From: SANTIA: a Matlab-based open-source toolbox for artifact detection and removal from extracellular neuronal signals

Network Sequence length (ms) Training loss Val. Acc. Test Acc. Test AUROC
MLP 50 0.20 0.92 0.92 0.98
100 0.41 0.82 0.81 0.90
150 0.39 0.83 0.83 0.90
200 0.24 0.91 0.91 0.97
1D-CNN 50 0.10 0.96 0.97 0.99
100 0.39 0.84 0.84 0.89
150 0.37 0.83 0.83 0.91
200 0.36 0.83 0.83 0.91
LSTM 50 0.16 0.93 0.94 0.99
100 0.26 0.90 0.91 0.97
150 0.25 0.89 0.90 0.97
200 0.25 0.91 0.90 0.97
  1. Values pertaining to model’s best performance are highlighted in bold