Struct DataSet 

pub struct DataSet {
    pub x: Array2<f64>,
    pub y: Array1<f64>,
    pub feature_names: Vec<String>,
    pub target_name: String,
}

A dataset of input features x (shape [n_rows, n_vars]) and targets y ([n_rows]).

Fields

x: Array2<f64>

Feature matrix, one row per observation.

y: Array1<f64>

Target vector, one entry per observation.

feature_names: Vec<String>

Names of the feature columns (length n_vars).

target_name: String

Name of the target column.

Implementations

impl DataSet

pub fn from_arrays(x: Array2<f64>, y: Array1<f64>) -> Result<Self>

Build a dataset from in-memory arrays.

Errors

Returns PhopError::ShapeMismatch if the row counts of x and y differ.

pub fn n_vars(&self) -> usize

Number of feature variables.

pub fn len(&self) -> usize

Number of observations.

pub fn is_empty(&self) -> bool

Whether the dataset is empty.

pub fn standardized(&self) -> (DataSet, Standardizer)

Produce a z-scored copy of the dataset together with the Standardizer that maps predictions back to the original target units.

Each feature column and the target are centered and scaled to unit variance; constant columns (zero variance) are left centered with a unit scale so the transform stays finite.

pub fn select(&self, rows: &[usize]) -> Result<DataSet>

Build a sub-dataset from the given row indices (used by minibatching).

Errors

Returns PhopError::ShapeMismatch if any index is out of range.

pub fn minibatches(&self, size: usize, seed: u64) -> Vec<DataSet>

Partition the data axis into shuffled minibatches of (at most) size rows.

The shuffle is seeded for reproducibility (Risk T3 mitigation: bounds per-step memory by letting the optimizer consume the data in chunks). A size of 0 or one >= the row count yields a single batch containing all rows.

pub fn from_csv<P: AsRef<Path>>(path: P) -> Result<Self>

Load a dataset from a CSV file.

The file is expected to have a header row. By default the last column is taken as the target y and all preceding columns as features x.

Errors

Returns an error if the file cannot be read, parsed, or has fewer than two columns.

pub fn from_csv_with_target<P: AsRef<Path>>( path: P, target: Option<usize>, ) -> Result<Self>

Load a dataset from a CSV file, optionally choosing which column is the target.

target is a 0-based column index; None selects the last column. All other columns become features x, preserving their header order.

Errors

Returns an error if the file cannot be read or parsed, has fewer than two columns, or if target is out of range.

pub fn from_csv_columns<P: AsRef<Path>>( path: P, features: &[usize], target: usize, ) -> Result<Self>

Load a dataset from a CSV file selecting an explicit subset of feature columns and a target column (all 0-based indices). Feature columns appear in the order given.

Errors

Returns an error if the file cannot be read/parsed, has fewer than two columns, any index is out of range, or the target appears among the features.

impl DataSet

pub fn to_dimensionless( &self, feature_dims: &[Dimension], ) -> Result<(DataSet, Vec<Vec<i32>>)>

Reduce the features to their dimensionless Buckingham-π groups, given each feature’s Dimension. The new feature columns are the monomials ∏ xᵢ^{eᵢ} for each π-group; the target is left unchanged. The π-group exponent vectors are returned alongside.

Errors

Returns PhopError::ShapeMismatch if feature_dims.len() != n_vars, or PhopError::NotConverged if the inputs are dimensionally independent (no π-groups exist).

Trait Implementations

impl Clone for DataSet

fn clone(&self) -> DataSet

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DataSet

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

Formats the value using the given formatter.