API Reference - WeightBlocks - Linear
Constructors
new()
Creates a new weight block object. If there are no parameters given for that particular argument, then that argument will use default value.
Linear.new({dimensionArray: {number}, learningRate: number, Optimizer: OptimizerObject, Regularizer: RegularizerObject, initializationMode: string}): WeightBlockObject
Parameters:
-
dimensionSizeArray: The dimensions for the weights. The length of array represents the number of dimensions. The value at the specified index represents the size at that dimension.
-
learningRate: The speed at which the model learns. Recommended that the value is set between (0 to 1).
-
Optimizer: The optimizer to be used.
-
Regularizer: The regularizer to be used.
-
initializationMode: The mode for the weights to be initialized. Available options are:
-
Zero
-
Random
-
RandomNormal
-
RandomUniformPositive
-
RandomUniformNegative
-
RandomUniformNegativeAndPositive
-
HeUniform
-
HeNormal
-
XavierUniform
-
XavierNormal
-
LeCunUniform
-
LeCunNormal
-
None
-
Returns:
- WeightBlock: The generated weight block object.