Kalax Python Interface

Python bindings for Kalax, a high-performance time series feature extraction library written in Rust.

Overview

Kalax provides fast, parallelized extraction of statistical features from time series data. The Python interface offers the same functionality as the Rust core with minimal overhead.

Installation

[uv] pip install kalax

Quick Start

import numpy as np
from kalax import extract_features

x = np.linspace(0, 1, 100)
y = np.sin(2 * np.pi * 5 * x) + 0.5
data = [{"x": x, "y": y}]
features = extract_features(data)

API

extract_features(data)

Extracts features from multiple time series in parallel.

Parameters:

• data (List[Dict[str, np.ndarray]]): A list of dictionaries where each dictionary maps column names to 1D numpy arrays representing time series data.

Returns:

• List[Dict[str, Dict[str, float]]]: A list of dictionaries where each entry corresponds to an input dictionary. Feature names are mapped to dictionaries containing computed feature values.

Available Features

Example

import numpy as np
from kalax import extract_features

x = np.linspace(0, 1, 100)
y = np.sin(2 * np.pi * 5 * x) + 0.5
data = [{"x": x, "y": y}]
features = extract_features(data)

print(features[0]["x"]["mean"])       # Access specific feature
print(features[0]["y"]["variance"])   # Access another feature

Output Structure

[
    {
        'column_name': {
            'mean': 0.5,
            'median': 0.5,
            'variance': 0.085,
            'standard_deviation': 0.292,
            'minimum': 0.0,
            'maximum': 1.0,
            'absolute_maximum': 1.0,
            'root_mean_square': 0.579,
            'sum_values': 50.0,
            'length': 100.0
        }
    }
]

Requirements

• Python 3.12+
• numpy

License

MIT