Table 3 Hyperparameters of ML classifiers.
From: SpectroFusionNet a CNN approach utilizing spectrogram fusion for electric guitar play recognition
Model category | Hyperparameter | Possible settings | Best setting |
|---|---|---|---|
Random forest | Number of trees | 10, 50, 100, 200 | 200 |
Maximum depth | None, 10, 20, 30 | None | |
Bootstrap | True, false | True | |
SVM | Kernel | Linear, RBF | RBF |
C (regularization) | 0.1, 1, 10 | 10 | |
Gamma | Scale, auto | Scale | |
k-NN | Number of neighbors | 3, 5, 7 | 3 |
Weights | Uniform, Distance | Distance | |
LMT | Class weights | 1:1, 1:2, 1:3 | 1:1 |
Naive Bayes | Var smoothing | 1e − 09, 1e − 08 | 1e − 09 |
MLP | Hidden layer sizes | (50,), (100,) | 100 |
Activation | ReLU, Tanh | ReLU | |
Solver | LBFGS, Adam | Adam | |
Decision tree | Criterion | Gini, Entropy | Gini |
Maximum depth | None, 10, 20 | None | |
Gradient booster | Number of estimators | 50, 100 | 100 |
Learning rate | 0.01, 0.1 | 0.1 | |
AdaBoost | Number of estimators | 50, 100 | 100 |
Algorithm | SAMME, SAMME.R | SAMME.R |
