Struct LogNormalVI 

pub struct LogNormalVI {
    pub dim: usize,
    pub mu: Array1<f64>,
    pub log_sigma: Array1<f64>,
}

Variational family using a log-normal distribution for positive-valued variables

z ~ LogNormal(mu, sigma^2), so log(z) ~ N(mu, sigma^2). This is suitable as a variational family for positive-valued latent variables.

Parameters (stored in flat form): [mu_0, …, mu_{d-1}, log_sigma_0, …, log_sigma_{d-1}]

Reparameterization: z_i = exp(mu_i + sigma_i * epsilon_i)

Fields

dim: usize

Dimension of the latent variable

mu: Array1<f64>

Mean parameters of log z: mu_i

log_sigma: Array1<f64>

Log standard deviation parameters: log_sigma_i (sigma_i = exp(log_sigma_i))

Implementations

impl LogNormalVI

pub fn new(dim: usize) -> Result<Self>

Create a new LogNormal variational family

Initializes with mu=0 and log_sigma=0 (sigma=1, so lognormal(0,1)).

pub fn from_params(mu: Array1<f64>, log_sigma: Array1<f64>) -> Result<Self>

Create from explicit parameters

pub fn sigmas(&self) -> Array1<f64>

Get the standard deviations sigma_i = exp(log_sigma_i)

pub fn means(&self) -> Array1<f64>

Compute the log-normal mean: E[z_i] = exp(mu_i + sigma_i^2 / 2)

pub fn variances(&self) -> Array1<f64>

Compute the log-normal variance: Var[z_i] = (exp(sigma_i^2) - 1) * exp(2*mu_i + sigma_i^2)

pub fn medians(&self) -> Array1<f64>

Compute the median: median[z_i] = exp(mu_i)

Trait Implementations

impl Clone for LogNormalVI

fn clone(&self) -> LogNormalVI

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LogNormalVI

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

Formats the value using the given formatter.

impl VariationalFamily for LogNormalVI

fn dim(&self) -> usize

Dimensionality of the latent variable z

fn n_params(&self) -> usize

Number of variational parameters

fn get_params(&self) -> Array1<f64>

Get current variational parameters as a flat vector

fn set_params(&mut self, params: &Array1<f64>) -> Result<()>

Set variational parameters from a flat vector

fn sample_reparam(&self, epsilon: &Array1<f64>) -> Result<(Array1<f64>, f64)>

Sample using the reparameterization trick: z = f(epsilon; phi), where epsilon ~ base_noise_distribution

fn entropy(&self) -> f64

Compute the differential entropy H[q] = -E_q[log q(z)]

fn log_prob(&self, z: &Array1<f64>) -> Result<f64>

Compute log q(z; phi) for a given sample z

fn reparam_gradient( &self, dlog_joint_dz: &Array1<f64>, epsilon: &Array1<f64>, ) -> Result<Array1<f64>>

Compute the gradient of the ELBO w.r.t. parameters using reparameterization.

fn kl_from_prior(&self) -> Option<f64>

Compute KL divergence KL(q || p) where p is a standard prior.