Struct ppca::PPCAModel

pub struct PPCAModel(_);

A PPCA model which can infer missing values.

Eeach sample for this model behaves according to the following statistical latent variable model.

x ~ N(0; I(nxn))
y = C * x + y0 + noise
noise ~ N(0; sgima ^ 2 * I(mxm))

Here, x is the latent state, y is the observed sample, that is an affine transformation of the hidden state contaminated by isotropic noise.

Note

All arrays involved have to be of data type float64.

Implementations

impl PPCAModel

pub fn new(
    isotropic_noise: f64,
    transform: DMatrix<f64>,
    mean: DVector<f64>
) -> PPCAModel

pub fn init(state_size: usize, dataset: &Dataset) -> PPCAModel

Creates a new random untrained model from a given latent state size and a dataset.

pub fn mean(&self) -> &DVector<f64>

Then center of mass of the distribution in the output space.

pub fn isotropic_noise(&self) -> f64

The standard deviation of the noise in the output space.

pub fn transform(&self) -> &DMatrix<f64>

The linear transformation from hidden state space to output space.

pub fn output_size(&self) -> usize

The number of features for this model.

pub fn state_size(&self) -> usize

The number of hidden values for this model.

pub fn n_parameters(&self) -> usize

The total number of parameters involved in training (used for information criteria).

pub fn singular_values(&self) -> DVector<f64>

The relative strength of each hidden variable on the output. This is equivalent to the eigenvalues in the standard PCA.

pub fn llk_one(&self, sample: &MaskedSample) -> f64

Compute the log-likelihood of a single sample.

pub fn llk(&self, dataset: &Dataset) -> f64

Compute the log-likelihood of a given dataset.

pub fn llks(&self, dataset: &Dataset) -> DVector<f64>

Compute the log-likelihood for each sample in a given dataset.

pub fn sample_one(&self, mask_prob: f64) -> MaskedSample

Sample a single sample from the PPCA model and masks each entry according to a Bernoulli (coin-toss) distribution of proability mask_prob of erasing the generated value.

pub fn sample(&self, dataset_size: usize, mask_prob: f64) -> Dataset

Sample a full dataset from the PPCA model and masks each entry according to a Bernoulli (coin-toss) distribution of proability mask_prob of erasing the generated value.

pub fn infer_one(&self, sample: &MaskedSample) -> InferredMasked

Infers the hidden components for one single sample. Use this method for fine-grain control on the properties you want to extract from the model.

pub fn infer(&self, dataset: &Dataset) -> Vec<InferredMasked>

Infers the hidden components for each sample in the dataset. Use this method for fine-grain control on the properties you want to extract from the model.

pub fn smooth_one(&self, sample: &MaskedSample) -> MaskedSample

Filters a single samples, removing noise from the extant samples and inferring the missing samples.

pub fn smooth(&self, dataset: &Dataset) -> Dataset

Filters each sample of a given datatset, removing noise from the extant samples and inferring the missing samples.

pub fn extrapolate_one(&self, sample: &MaskedSample) -> MaskedSample

Extrapolates the missing values with the most probable values for a single sample. This method does not alter the extant values.

pub fn extrapolate(&self, dataset: &Dataset) -> Dataset

Extrapolates the missing values with the most probable values for a full dataset. This method does not alter the extant values.

pub fn iterate(&self, dataset: &Dataset) -> PPCAModel

Makes one iteration of the EM algorithm for the PPCA over an observed dataset, returning the improved model. The log-likelihood will always increase for the returned model.

pub fn iterate_with_prior(&self, dataset: &Dataset, prior: &Prior) -> PPCAModel

Makes one iteration of the EM algorithm for the PPCA over an observed dataset, using a supplied PPCA prior and returning the improved model. This method will not necessarily increase the log-likelihood of the returned model, but it will return an improved maximum a posteriori (MAP) estimate of the PPCA model according to the supplied prior.

pub fn to_canonical(&self) -> PPCAModel

Returns a canonical version of this model. This does not alter the log-probablility function nor the quality of the training. All it does is to transform the hidden variables.

Trait Implementations

impl Clone for PPCAModel

fn clone(&self) -> PPCAModel

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for PPCAModel

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for PPCAModel

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>where
    __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for PPCAModel

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>where
    __S: Serializer,

Serialize this value into the given Serde serializer.