Struct DiffableGaussian2D 

pub struct DiffableGaussian2D<T: Float> {
    pub mean: [T; 2],
    pub cov: [[T; 2]; 2],
    pub inv_cov: [[T; 2]; 2],
    pub logdet_cov: T,
    pub norm_const: T,
}

A 2D Gaussian target distribution, parameterized by mean and covariance.

This struct also precomputes the inverse covariance and a log-normalization constant so we can quickly compute log-densities and gradients in unnorm_logp.

Fields

mean: [T; 2]

cov: [[T; 2]; 2]

inv_cov: [[T; 2]; 2]

logdet_cov: T

norm_const: T

Implementations

impl<T> DiffableGaussian2D<T>

where T: Float + Debug + FloatConst,

pub fn new(mean: [T; 2], cov: [[T; 2]; 2]) -> Self

Create a new 2D Gaussian with the specified mean and covariance. We automatically compute the covariance inverse and log-determinant.

Trait Implementations

impl<T, B> BatchedGradientTarget<T, B> for DiffableGaussian2D<T>

where T: Float + ElementConversion + Debug + Element, B: AutodiffBackend,

fn unnorm_logp_batch(&self, positions: Tensor<B, 2>) -> Tensor<B, 1>

Evaluate the log probability for a batch of positions: shape [n_chains, 2]. Return shape [n_chains]. Note: It is not necessary to return the log probability here but for easier debugging we do so anyways.

impl<T: Clone + Float> Clone for DiffableGaussian2D<T>

fn clone(&self) -> DiffableGaussian2D<T>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<T: Debug + Float> Debug for DiffableGaussian2D<T>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<T, B> GradientTarget<T, B> for DiffableGaussian2D<T>

where T: Float + ElementConversion + Debug + Element, B: AutodiffBackend,

fn unnorm_logp(&self, position: Tensor<B, 1>) -> Tensor<B, 1>

fn unnorm_logp_and_grad( &self, position: Tensor<B, 1>, ) -> (Tensor<B, 1>, Tensor<B, 1>)