Struct ArrayRng 

pub struct ArrayRng<A> { /* private fields */ }

Random number generator for ndarray

Uses ChaCha8Rng internally because StdRng does not implement Clone in rand 0.10.

Implementations

impl<A: Float> ArrayRng<A>

pub fn new() -> Self

Creates a new random number generator with the default seed.

pub fn from_seed(seed: u64) -> Self

Creates a new random number generator with the specified seed.

pub fn as_rng(&self) -> &ChaCha8Rng

Returns a reference to the internal RNG

pub fn as_rng_mut(&mut self) -> &mut ChaCha8Rng

Returns a mutable reference to the internal RNG

pub fn random(&mut self, shape: &[usize]) -> NdArray<A>

Creates a uniform random array in the specified shape. Values are in the range [0, 1)

pub fn normal(&mut self, shape: &[usize], mean: f64, std: f64) -> NdArray<A>

Creates a normal random array in the specified shape. Values are drawn from a normal distribution with the specified mean and standard deviation.

pub fn uniform(&mut self, shape: &[usize], low: f64, high: f64) -> NdArray<A>

Creates a uniform random array in the specified shape. Values are in the range [low, high).

pub fn glorot_uniform(&mut self, shape: &[usize]) -> NdArray<A>

Creates a random array with Glorot/Xavier uniform initialization. For a tensor with shape (in_features, out_features), samples are drawn from Uniform(-sqrt(6/(in_features+out_features)), sqrt(6/(in_features+out_features))).

pub fn glorot_normal(&mut self, shape: &[usize]) -> NdArray<A>

Creates a random array with Glorot/Xavier normal initialization. For a tensor with shape (in_features, out_features), samples are drawn from Normal(0, sqrt(2/(in_features+out_features))).

pub fn he_uniform(&mut self, shape: &[usize]) -> NdArray<A>

Creates a random array with He/Kaiming uniform initialization. For a tensor with shape (in_features, out_features), samples are drawn from Uniform(-sqrt(6/in_features), sqrt(6/in_features)).

pub fn he_normal(&mut self, shape: &[usize]) -> NdArray<A>

Creates a random array with He/Kaiming normal initialization. For a tensor with shape (in_features, out_features), samples are drawn from Normal(0, sqrt(2/in_features)).

pub fn standard_normal(&mut self, shape: &[usize]) -> NdArray<A>

Creates a random array from the standard normal distribution.

pub fn standard_uniform(&mut self, shape: &[usize]) -> NdArray<A>

Creates a random array from the standard uniform distribution.

pub fn bernoulli(&mut self, shape: &[usize], p: f64) -> NdArray<A>

Creates a random array from the bernoulli distribution.

pub fn exponential(&mut self, shape: &[usize], lambda: f64) -> NdArray<A>

Creates a random array from the exponential distribution.

pub fn log_normal( &mut self, shape: &[usize], mean: f64, stddev: f64, ) -> NdArray<A>

Creates a random array from the log-normal distribution.

pub fn gamma( &mut self, shape: &[usize], shape_param: f64, scale: f64, ) -> NdArray<A>

Creates a random array from the gamma distribution.

Trait Implementations

impl<A: Clone> Clone for ArrayRng<A>

fn clone(&self) -> ArrayRng<A>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<A: Float> Default for ArrayRng<A>

fn default() -> Self

Returns the “default value” for a type.

impl<A> TryRng for ArrayRng<A>

type Error = Infallible

The type returned in the event of a RNG error.

fn try_next_u32(&mut self) -> Result<u32, Self::Error>

Return the next random u32.

fn try_next_u64(&mut self) -> Result<u64, Self::Error>

Return the next random u64.

fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), Self::Error>

Fill dst entirely with random data.