Topf — Topological peak filtering

topf is a small library for Python 3 that permits the detection and subsequent filtering of peaks in one-dimensional functions. The method is based on a topological notion of prominence or persistence of a peak with respect to all other peaks.

Dependencies

• Python 3.7
• numpy

Installation

• Clone the repository
• poetry install (for local development)
• pip3 install . (for global usage)

Usage

Install the library using pip3 install topf. You can then access the main class, PersistenceTransformer by issuing import topf. As a simple example, we load the file example.txt, depict its peaks, and filter the smallest ones:

import topf

import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns


data = np.genfromtxt('example.txt')          # load data
transformer = topf.PersistenceTransformer()  # prepare transformer
peaks = transformer.fit_transform(data)      # transform data into peaks

# First, let's plot the original data. We can see that there is
# quite a number of relatively small peaks.
plt.subplot(3, 1, 1)
sns.lineplot(x=data[:, 0], y=data[:, 1])

# Second, let's show the transformed data. Here, every non-zero
# point depicts the *prominence* of a peak.
plt.subplot(3, 1, 2)
sns.lineplot(x=peaks[:, 0], y=peaks[:, 1])

# Third, let's show an example of filtering. At present, there is
# no automated way of doing so.
filtered_data = data[peaks[:, 1] > 4]  # only keep high peaks

plt.subplot(3, 1, 3)
sns.lineplot(x=data[:, 0], y=data[:, 1], alpha=0.5)
sns.scatterplot(
    x=filtered_data[:, 0],
    y=filtered_data[:, 1],
)

plt.tight_layout()
plt.show()

This file is also available as example.py in this repository (with some minor modifications to simplify usage). It will result in the following output:

Licence notice

The icon of this project was created by Freepik from www.flaticon.com and is licensed by CC 3.0 BY.