setHyperParams: Hyper-parameter setter

In ff98li/scRGNet: Perform scRNA-seq Analysis using Single-cell Graphical Neural Network (scGNN) Framework

Description

A function to set up the hyper-parameter to tune the auto-encoder

Usage

setHyperParams(
  batch_size = 1L,
  regu_epochs = 5L,
  L1 = 0.5,
  L2 = 0.5,
  regu_alpha = 0.9,
  reduction = "sum"
)

Arguments

batch_size	Batch size used to train the auto-encoder
regu_epochs	Epoch of trainning; that is, the number of time the dataset is visited.
L1	Intensity of L1 penalty on loss if LTMG matrix is used. Takes value in range >=0 and >= 1
L2	Intensity of L2 penalty on loss if LTMG matrix is used. Takes value in range >=0 and >= 1
regu_alpha	Intensity of LTMG regularisation. Takes value in range >=0 and >= 1
reduction	Type of reduction used in loss functions: available options: "mean", "sum", or "none", meaning not using reduction method.

Value

A list of hyper-parameters for runFeatureAE

References

\insertRef

scGNNscRGNet

ff98li/scRGNet documentation built on Jan. 14, 2022, 4:58 a.m.