Class GAN

constructor

• new GAN(
      noise_dim: number,
      g_hidden: LayerObject[],
      d_hidden: LayerObject[],
      g_opt: string,
      d_opt: string,
      class_size: number,
      type: "" | "conditional",
  ): GAN

  Parameters

  • noise_dim: number

    Number of noise dimension
  • g_hidden: LayerObject[]

    Layers of generator
  • d_hidden: LayerObject[]

    Layers of discriminator
  • g_opt: string

    Optimizer of the generator network
  • d_opt: string

    Optimizer of the discriminator network
  • class_size: number

    Class size for conditional type
  • type: "" | "conditional"

    Type name

  Returns GAN

_discriminator

_epoch

_g_opt

_generator

_generatorNetLeyers

_noise_dim

_type

epoch

• get epoch(): number

  Epoch

  Returns number

fit

• fit(
      x: number[][],
      y: number[][],
      step: number,
      gen_rate: number,
      dis_rate: number,
      batch: number,
  ): { discriminatorLoss: number; generatorLoss: number }

  Fit model.

  Parameters

  • x: number[][]

    Training data
  • y: number[][]

    Conditional values
  • step: number

    Iteration count
  • gen_rate: number

    Learning rate for generator
  • dis_rate: number

    Learning rate for discriminator
  • batch: number

    Batch size

  Returns { discriminatorLoss: number; generatorLoss: number }

  Loss value

generate

• generate(n: number, y: number[][]): number[][]

  Returns generated data from the model.

  Parameters

  • n: number

    Number of generated data
  • y: number[][]

    Conditional values

  Returns number[][]

  Generated values

prob

• prob(x: number[][], y: any): number[][]

  Returns probabilities of the data is true.

  Parameters

  • x: number[][]

    Sample data
  • y: any

    Conditional values

  Returns number[][]

  Predicted values