A laminar augmented cascading flexible neural forest model for classification of cancer subtypes based on gene expression data

Table 1 Summarizes the hyper-parameters of deep neural networks (e.g., convolutional neural networks) and LACFNForest, where the default values used in our experiments are given

Parameter of LACFNForest	Value	Hyper-parameter of CNN	Value
Population size	50	No. hidden layers	Uncertain
Crossover probability	0.4	No. feature maps	Uncertain
Mutation probability	0.01	No. nodes in hidden layers	Uncertain
Level	4	Learning rate	Uncertain
C1	2.0	Kernal size	Uncertain
C2	2.0	Momentum	Uncertain
Vmax	2.0	L1/L2 weight regularization penalty	Uncertain

“Uncertain” indicates that the default values of the parameters are unknown or usually require different settings for different tasks. These hyperparameters do not exist in our model, or exist but can be determined automatically during the optimization process of our model

ISSN: 1471-2105