Struct TimeSeriesVAE 

pub struct TimeSeriesVAE<F: Float + Debug + ScalarOperand> { /* private fields */ }

Variational Autoencoder for Time Series with Uncertainty Quantification

Implementations

impl<F: Float + Debug + Clone + FromPrimitive + ScalarOperand> TimeSeriesVAE<F>

pub fn new( seq_len: usize, feature_dim: usize, latent_dim: usize, encoder_hidden: usize, decoder_hidden: usize, ) -> Self

Create new Time Series VAE

pub fn encode(&self, input: &Array2<F>) -> Result<(Array1<F>, Array1<F>)>

Encode time series to latent distribution

pub fn reparameterize(&self, mean: &Array1<F>, logvar: &Array1<F>) -> Array1<F>

Sample from latent distribution using reparameterization trick

pub fn decode(&self, latent: &Array1<F>) -> Result<Array2<F>>

Decode latent representation to time series

pub fn forward(&self, input: &Array2<F>) -> Result<VAEOutput<F>>

Full forward pass with reconstruction and KL divergence

pub fn generate(&self, numsamples: usize) -> Result<Vec<Array2<F>>>

Generate new time series by sampling from latent space

pub fn estimate_uncertainty( &self, input: &Array2<F>, num_samples: usize, ) -> Result<(Array2<F>, Array2<F>)>

Estimate uncertainty by sampling multiple reconstructions

Trait Implementations

impl<F: Debug + Float + Debug + ScalarOperand> Debug for TimeSeriesVAE<F>

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

Formats the value using the given formatter.