Basic Example

This example shows the data path from CSV ingestion to tensor training.

Objective

Demonstrate the numeric ingestion path with minimal assumptions about input schema.

Data flow

CSV file
   ├─ numeric-only fast path -> Tensor mode
   └─ mixed-type fallback -> ETL mode -> materialize -> Tensor mode

Example

use Pml\Dataset;
use Pml\Estimators\Regression\GBDTRegressor;

$dataset = Dataset::fromCSV('datasets/housing/train.csv', labelColumn: 0);
$model = new GBDTRegressor();
$model->train($dataset);

Internals

• Dataset::fromCSV() attempts tensor_dataset_from_csv() first.
• If the CSV contains mixed types, the framework falls back to df_read_csv().
• materialize() converts the DataFrame to native tensors once.

Performance notes

• Numeric-only ingestion avoids ETL overhead.
• If the CSV contains categorical columns, prefer ETL transformations before materialize().