VarianceScaling

open class VarianceScaling(scale: Double, mode: Mode, distribution: Distribution, seed: Long) : Initializer

Initializer capable of adapting its scale to the shape of weights tensors.

With distribution="truncated_normal" or "untruncated_normal", samples are drawn from a truncated/untruncated normal distribution with a mean of zero and a standard deviation (after truncation, if used) stddev = sqrt(scale / n) where n is:

number of input units in the weight tensor, if mode = "fan_in"
number of output units, if mode = "fan_out"
average of the numbers of input and output units, if mode = "fan_avg"

With distribution="uniform", samples are drawn from a uniform distribution within -limit, limit, with limit = sqrt(3 * scale / n).

Constructors

VarianceScaling

fun VarianceScaling(scale: Double = 1.0, mode: Mode = Mode.FAN_IN, distribution: Distribution = Distribution.TRUNCATED_NORMAL, seed: Long = 12L)

Creates VarianceScaling initializer.

Functions

apply

fun apply(fanIn: Int, fanOut: Int, tf: Ops, input: Operand<Float>, name: String): InitializerOperation

Adds an Assign Op to the graph to initialize a tensorflow variable as specified by the initializer.

initialize

open override fun initialize(fanIn: Int, fanOut: Int, tf: Ops, shape: Operand<Int>, name: String): Operand<Float>

Returns a Tensor object initialized as specified by the initializer.

toString

open override fun toString(): String

Properties

distribution

val distribution: Distribution

Random distribution to use. One of "truncated_normal", "untruncated_normal" and "uniform".

mode

val mode: Mode

One of "fan_in", "fan_out", "fan_avg".

scale

val scale: Double = 1.0

Scaling factor (should be positive).

seed

val seed: Long = 12L

Used to create random seeds.

Inheritors

GlorotNormal

GlorotUniform

HeNormal

HeUniform

LeCunNormal

LeCunUniform