Options
All
  • Public
  • Public/Protected
  • All
Menu

Class StepRate

Step Learning Rate

The learning rate will decrease (i.e. 'step down') every stepSize iterations.

Hierarchy

Index

Constructors

Properties

Methods

Constructors

constructor

  • new StepRate(baseRate: number, gamma?: number, stepSize?: number): StepRate

  • Constructs a step rate policy.

    Parameters

    • baseRate: number

      the rate at first iteration

    • Default value gamma: number = 0.9
    • Default value stepSize: number = 100

    Returns StepRate

Properties

Protected Readonly baseRate

baseRate: number

The rate at the first iteration.

Private Readonly gamma

gamma: number

Learning rate retention per step; - 0 < gamma < 1 - large gamma CAN cause networks to never converge, low gamma CAN cause networks to converge too quickly

Private Readonly stepSize

stepSize: number

Learning rate is updated every step_size iterations

Methods

calc

  • calc(iteration: number): number

  • Calculates the current training rate.

    Parameters

    • iteration: number

      count

    Returns number

    the current training rate

Generated using TypeDoc