muffler 0.2.0

Denoising time-series data.

Muffler

Denoising Time-Series Data

Installation

pip install muffler

Usage

import muffler
import numpy as np

# Generate some noisy data
t = np.linspace(0, 1, 100)
clean_signal = np.sin(2 * np.pi * t)

# Add some noise
std = 1.0
conditional_noise = np.sqrt(np.abs(clean))
random_noise = np.random.normal(0, std, t.shape)
noisy_signal = clean + random_noise * conditional_noise

# Denoise the data
window_size = 100
stride = 25
denoised_signal = muffler.denoise_linear_regression(noisy_signal, window_size, stride)

# Compute the error
error = np.mean(np.sqrt(np.mean((clean_signal - denoised_signal) ** 2, axis=1)))
print(f"Error: {error:.2e}")

Streamlit App (An example)

You can also use the streamlit app to see how the denoising works. To run the app, you need to install Rust and Maturin for compiling the code. Then, you need to create a Python (3.9 or greater) virtual environment and install the dependencies:

python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
maturin develop --release --extras=dev

Finally, you can run the app:

streamlit run examples/app.py

License

This project is licensed under the MIT License - see the LICENSE file for details.

Citation

TODO: Add citation