Show / Hide Table of Contents

Class MinMaxNorm

MinMaxNorm weight constraint. Constrains the weights incident to each hidden unit to have the norm between a lower bound and an upper bound.

Inheritance
System.Object
BaseConstraint
MinMaxNorm
Namespace: SiaNet.Constraints
Assembly: SiaNet.dll
Syntax
public class MinMaxNorm : BaseConstraint

Constructors

| Improve this Doc View Source

MinMaxNorm(Single, Single, Single, UInt32)

Initializes a new instance of the MinMaxNorm class.

Declaration
public MinMaxNorm(float minVale = 0F, float maxValue = 1F, float rate = 1F, uint axis = 0U)
Parameters
Type Name Description
System.Single minVale

The minimum norm for the incoming weights.

System.Single maxValue

The maximum norm for the incoming weights.

System.Single rate

Rate for enforcing the constraint: weights will be rescaled to yield (1 - rate) * norm + rate * norm.clip(min_value, max_value).

System.UInt32 axis

Integer, axis along which to calculate weight norms.

Properties

| Improve this Doc View Source

Axis

Integer, axis along which to calculate weight norms. For instance, in a Dense layer the weight matrix has shape (input_dim, output_dim), set axis to 0 to constrain each weight vector of length (input_dim,). In a Conv2D layer, the weight tensor has shape (output_depth, input_depth, rows, cols), set axis to [1, 2, 3] to constrain the weights of each filter tensor of size (input_depth, rows, cols).

Declaration
public uint Axis { get; set; }
Property Value
Type Description
System.UInt32

The axis.

| Improve this Doc View Source

MaxValue

The maximum norm for the incoming weights.

Declaration
public float MaxValue { get; set; }
Property Value
Type Description
System.Single

The maximum value.

| Improve this Doc View Source

MinValue

The minimum norm for the incoming weights.

Declaration
public float MinValue { get; set; }
Property Value
Type Description
System.Single

The minimum value.

| Improve this Doc View Source

Rate

Rate for enforcing the constraint: weights will be rescaled to yield (1 - rate) * norm + rate * norm.clip(min_value, max_value). Effectively, this means that rate=1.0 stands for strict enforcement of the constraint, while rate<1.0 means that weights will be rescaled at each step to slowly move towards a value inside the desired interval.

Declaration
public float Rate { get; set; }
Property Value
Type Description
System.Single

The rate.

See Also

BaseConstraint
  • Improve this Doc
  • View Source
Back to top Generated by DocFX