DataTableNorm

A normalizer for DataTable objects that handles feature scaling, supporting both Min-Max and Z-Score normalization.

Overview

The DataTableNorm class provides feature normalization capabilities for machine learning datasets stored in DataTable format. It supports two normalization techniques:

• Min-Max Scaling: Transforms features to [0, 1] range
• Z-Score Standardization: Transforms features to mean=0, std=1

Key Features:

• Computes parameters from training data
• Handles both in-place and copied normalization
• Supports efficient rvalue operations
• Works with evaluation/test splits

---

Constructor

DataTableNorm(const DataTable<T> &data, NormalizationType type = MIN_MAX)

Initializes the normalizer from a data table.

Example:

txeo::DataTable<double> data = load_my_dataset();
txeo::DataTableNorm<double> normalizer(data, txeo::NormalizationType::Z_SCORE);

---

Member Functions

const DataTable<T>& data_table() const

Returns the internal reference to the associated DataTable.

Example:

const auto& dt = normalizer.data_table();
std::cout << dt.x_train().rows() << std::endl;

void set_data_table(const DataTable<T>& data)

Sets a new data table for normalization.

Example:

txeo::DataTable<double> new_data = load_updated_dataset();
normalizer.set_data_table(new_data);

NormalizationType type() const

Returns the type of normalization currently used.

Example:

if (normalizer.type() == txeo::NormalizationType::MIN_MAX)
    std::cout << "Using Min-Max normalization" << std::endl;

Matrix<T> normalize(Matrix<T>&& x) const

Normalizes a matrix in-place using rvalue semantics.

Example:

txeo::Matrix<double> large_matrix = generate_large_data();
auto normalized = normalizer.normalize(std::move(large_matrix));

Matrix<T> normalize(const Matrix<T>& x) const

Normalizes a matrix by copy.

Example:

txeo::Matrix<double> original = {{1.0}, {2.0}, {3.0}};
auto normalized = normalizer.normalize(original);

Matrix<T> x_train_normalized()

Returns normalized training data.

Example:

auto x_train_norm = normalizer.x_train_normalized();
model.train(x_train_norm, normalizer.data_table().y_train());

Matrix<T> x_eval_normalized()

Returns normalized evaluation data.

Example:

auto x_eval_norm = normalizer.x_eval_normalized();
model.evaluate(x_eval_norm, normalizer.data_table().y_eval());

Matrix<T> x_test_normalized()

Returns normalized test data.

Example:

auto x_test_norm = normalizer.x_test_normalized();
model.test(x_test_norm, normalizer.data_table().y_test());

Exceptions

• DataTableNormError: Thrown if normalization parameters are invalid or data table is inconsistent.

---

Notes

• Normalization parameters are computed from training data only.
• Be sure to properly configure your DataTable before using DataTableNorm.

For detailed API references, see individual method documentation at txeo::DataTableNorm.