pyholo 0.1.8

Multipurpose interpolation library focussing on RBF, written in rust.

PyHolo

Python bindings for flexible interpolation library featuring Radial Basis Function (RBF) written in rust. It features:

• RBF interpolation with many kernels to choose from.
• Flexible Data Types: Supports input and output datasets of type float, list[float], ndarray.

Installation

pip install pyholo

For optimal performance, it's recommended to use a virtual environment.

Quick Start

1D Interpolation of a target function

import numpy as np
import pyholo as holo

def target_function(pts):
    return 10.0 + np.sum(
        np.power(pts, 2) + 10.0 * np.cos(2.0 * np.pi * pts)
    )

# Get or set input/output data
x_train = np.linspace(-5.0, 5.0, 50)
y_train = np.array([target_function(x) for x in x_train])

# Get predictions from each model
pyholo_pred = holo.Rbf(x_train, y_train, "gaussian", 1.0)

# Points to predict at
x_new = np.linspace(-5.0, 5.0, 500)

# Make predictions
y_pred = pyholo_pred.predict(x_new)

Plotting against other Rbf implementations (scipy, numpy) and the expected values:

3D interpolation of data

import numpy as np
import pyholo as holo

# Training data
x_train = np.array([
    0.000, 512.000, 1182.490, 1911.273, 2788.547, 
    4227.750, 6481.706, 9609.367, 11773.210, 
    13188.649, 14400.000
])

y_train = np.array([
    [6950.000, 0.000, 0.000],
    [5930.999, 4386.866, 974.859],
    [2498.780, 8537.536, 1897.230],
    [-1993.234, 10819.437, 2404.319],
    [-7051.251, 11465.688, 2547.931],
    [-13549.361, 9748.529, 2166.340],
    [-19093.612, 3887.659, 863.924],
    [-18104.032, -5716.279, -1270.284],
    [-11425.229, -10664.783, -2369.952],
    [-4206.775, -11312.308, -2513.846],
    [3167.219, -7987.115, -1774.914],
])

# Create and fit RBF model
rbf = holo.Rbf(x_train, y_train, kernel='thin_plate_spline', epsilon=1.0)

# Predict on test points
x_test = np.linspace(x_train.min(), x_train.max(), 200)
y_pred = rbf.predict(x_test)

Plotting against the training data and expected values:

Comparison with other Rbf implementations

Using the hologram linear solver only, as haven't quite figured out how to configure Openblas or Mkl with Pyo3. Help here is appreciated. Simple 1d problem from before benchmarks quite well.

Build python from source

These instructions assume that Python3 and Cargo are installed on your system. To set up this project, follow these steps:

1. Clone the repository:

   git clone https://github.com/alexlovric/gmac.git
   cd gmac/gmac_py
   

2. Create a virtual environment and install build system:

   python3 -m venv .venv
   source .venv/bin/activate # In windows /Scripts/activate
   python3 -m pip install -r requirements.txt
   

3. Build the release binary:

   maturin develop --release
   

4. Build the python wheel:

   maturin build --release
   

5. Running benchmarks/examples:

   python3 -m pip install <path to wheel (target/wheels/*.whl)>
   cd benches
   python3 -m pip install -r requirements.txt
   

License

MIT License - See LICENSE for details.