KDEpy

1.1.12 verified Fri May 01 auth: no python

KDEpy is a Python library for kernel density estimation (KDE). It provides implementations of various KDE methods including naive, convolution, FFT, and tree-based approaches. Current version is 1.1.12. Release cadence is irregular, with updates primarily for compatibility and bug fixes.

pip install kdepy

Common errors

error ModuleNotFoundError: No module named 'kdepy' ↓

cause Incorrect import casing. The correct package name is 'KDEpy' with uppercase K, D, E.

fix

Install with pip install kdepy (lowercase) but import as from KDEpy import ....

error AttributeError: module 'KDEpy' has no attribute 'KDE' ↓

cause Attempting to access a class that doesn't exist. KDEpy's main class is `FFTKDE`, not `KDE`.

fix

Use from KDEpy import FFTKDE instead of from KDEpy import KDE.

error ValueError: The bandwidth parameter must be a positive float, a string, or a callable. ↓

cause Passing an invalid value for `bw`. Common mistake: passing an integer or negative number.

fix

Ensure bw is a positive float, one of the supported string methods ('silverman', 'SCOTT', 'ISJ'), or a callable.

error TypeError: fit() got multiple values for argument 'data' ↓

cause Calling `fit` with positional argument after keyword argument, e.g., `kde.fit(bw='ISJ', data)`. Fit expects first positional argument as data.

fix

Use kde.fit(data, bw='ISJ') or kde.fit(data) after setting bw earlier.

error ImportError: cannot import name 'KDE' from 'KDEpy' (unknown location) ↓

cause Old API. In earlier versions (<1.0) the main class was named `KDE`. It was renamed to `FFTKDE` in v1.0.

fix

Upgrade to latest version and use FFTKDE. If stuck on old version, import KDE.

Warnings

gotcha The import path uses mixed case: `from KDEpy import ...`. Lowercase `from kdepy import ...` will fail. ↓

fix Use uppercase KDE in the import statement.

gotcha The Bandwidth class is not exposed in the top-level namespace. Users may mistakenly attempt to import it directly from KDEpy. ↓

fix Use `from KDEpy.bandwidth import Bandwidth` or rely on automatic bandwidth selection via `bw` parameter.

gotcha In KDEpy, the default bandwidth selection method changed between versions. In v1.1.0+, the default for FFTKDE is 'ISJ' (Improved Sheather-Jones) instead of 'silverman'. ↓

fix Explicitly specify `bw='silverman'` if you need the old default.

deprecated The `bw_method` keyword argument was deprecated in v1.0.0 in favor of `bw`. ↓

fix Use `bw` instead of `bw_method`.

Imports

KDE
wrong
```
from kdepy import KDE
```
correct
```
from KDEpy import KDE
```
Package name uses uppercase KDE in import.
FFTKDE
```
from KDEpy import FFTKDE
```
NaiveKDE
```
from KDEpy import NaiveKDE
```
TreeKDE
```
from KDEpy import TreeKDE
```

Quickstart

Fit a kernel density estimate using FFT-based KDE with Improved Sheather-Jones bandwidth selection.

import numpy as np
from KDEpy import FFTKDE

# Generate sample data
np.random.seed(123)
data = np.random.normal(0, 1, size=100)

# Fit KDE
kde = FFTKDE(bw='ISJ')
x, y = kde.fit(data).evaluate(1024)

# Plot
import matplotlib.pyplot as plt
plt.plot(x, y)
plt.show()