README.md

References

Original Paper

Deep Ensemble Learning with Jet-Like architecture

1. Tail > Sparse Autoencoder
2. Wings > Two CNNs
3. Body > Non-Linear PCA
4. Nose > MLP

Architectural details

CNN1 (Left Wing)

conv layers = 3
fc layers = 1
filter size = (1,4)
input channels = 1
output channels = 3
stride = 1
padding = none
output features = 15

CNN2 (Right Wing)

conv layers = 4
fc layers = 1
filter size = (1,3)
input channels = 1
output channels = 3
stride = 1
padding = none
output features = 15

Sparse AE (Tail)

linear layers = 8
dropout = 0.1
sparsity = 0.00005
output features = 15

NLPCA (Body)

input features = 108
linear layers = 3
output features = 15

MLP (Nose)

linear layers = 3