gizai:ml/data

the data (row-major f64 matrices), the estimators and the model that the other parts share.

import * as Data from "gizai:ml/data";

Types

Data

type Data = { rows: i64; features: i64; x: Vec<f64>; targets: i64; y: Vec<f64>; }

A training set: rows × features in x, rows × targets in y (labels for classification, none for clustering).

• rows: i64
• features: i64
• x: Vec<f64>
• targets: i64
• y: Vec<f64>

Split

type Split = { train: Data; validation: Data; }

Two disjoint copies of one dataset, preserving each row's feature/target pairing.

• train: M37.Data
• validation: M37.Data

Functions

tabular

function tabular(features: Vec<Vec<f64>>, targets: Vec<Vec<f64>> = []): Data throws string

Build a row-major dataset without a database. Labels are optional for clustering or PCA. Every feature row and target row must have one consistent width.

Effects: fail.

param

function param(h: Json, key: string): Option<Json>

The hyperparameter key of h, if set.

Effects: fail.

number

function number(h: Json, key: string, d: f64): f64

The number under key in the hyperparameters h, else d.

Effects: fail.

integer

function integer(h: Json, key: string, d: i64): i64

The integer under key in the hyperparameters h, else d.

Effects: fail.

text

function text(h: Json, key: string, d: string): string

The text under key in the hyperparameters h, else d.

Effects: fail.

flag

function flag(h: Json, key: string, d: bool): bool

The boolean hyperparameter KEY of H, or D when H does not have it.

Effects: fail.

isNull

function isNull(h: Json, key: string): bool

Whether the hyperparameter key of h is set to null.

Effects: fail.

maxFeatures

function maxFeatures(h: Json, d: i64, fallback: string): i64

max_features: "sqrt", "log2", an int, a fraction, or null (all).

Effects: fail.

textual

function textual(v: Json): bool

Whether v is a JSON string.

column

function column(y: Vec<f64>, n: i64, t: i64, j: i64): Vec<f64>

Column j of the n by t row-major matrix y.

Effects: fail.

Methods

Data.validate

function Data.validate(self: Data, task: string): bool throws string

Validate the physical data contract before any algorithm indexes a row.

Effects: fail.

Data.split

function Data.split(self: Data, validationRows: i64, seed: u64): Split throws string

Deterministic shuffled holdout. Fit preprocessing on train, never on validation. For class-stratified or time-series splits, select rows in the caller instead.

Effects: fail.

Data.folds

function Data.folds(self: Data, k: i64, seed: u64): Vec<Split> throws string

K disjoint validation folds: each row appears in exactly one validation set. Each fold trains on the remaining rows; deterministic for a given seed.

Effects: fail.

Data.stratifiedFolds

function Data.stratifiedFolds(self: Data, k: i64, seed: u64): Vec<Split> throws string

Stratified K-folds for classification: preserve class proportions in every fold. Require at least K examples per class to keep minority classes evaluable.

Effects: fail.

Model.predictChecked

function Model.predictChecked(self: Model, row: Vec<f64>): Vec<f64> throws string

Safe single-row inference. Use this at untrusted API and file boundaries.

Effects: fail.

Model.predictRowsChecked

function Model.predictRowsChecked(self: Model, x: Vec<f64>, rows: i64): Vec<f64> throws string

Safe inference of a row-major batch without a database.

Effects: fail.

Model.probaChecked

function Model.probaChecked(self: Model, row: Vec<f64>): Vec<f64> throws string

Safe probability inference for one row.

Effects: fail.

Model.outputs

function Model.outputs(self: Model, row: Vec<f64>): Vec<f64>

The raw outputs of a row: the concatenation of the outputs of each part.

Effects: fail.

Model.predict

function Model.predict(self: Model, row: Vec<f64>): Vec<f64>

The prediction of a row: each target (regression), the label (classification), the cluster (clustering), the components (decomposition).

Effects: fail.

Model.predictRows

function Model.predictRows(self: Model, x: Vec<f64>, n: i64): Vec<f64>

The predictions of n rows of x (row-major, features wide), one row after another, on all threads.

Effects: fail.

Model.probaRows

function Model.probaRows(self: Model, x: Vec<f64>, n: i64): Vec<f64> throws string

The class probabilities of n rows of x, one row after another, on all threads.

Effects: fail.

Model.proba

function Model.proba(self: Model, row: Vec<f64>): Vec<f64> throws string

The probability of each class, in the order of the classes.

Effects: fail.

Constants

ML_FORMAT

const ML_FORMAT = 1

The version of the model file format.