scikit-misc

0.5.2 · active · verified Thu Apr 16

scikit-misc is a Python library providing a collection of miscellaneous tools for data analysis and scientific computing. It includes algorithms that were once part of SciPy's `scipy.misc` module but were removed or deemed unstable, now offered in a more robust form. The library is currently at version 0.5.2 and maintains an active development and release cadence, requiring Python >=3.10.

Common errors

```
ModuleNotFoundError: No module named 'scikit-misc'
```
cause The `scikit-misc` package is either not installed or not installed in the Python environment currently in use.

fix Run `pip install scikit-misc` or `conda install conda-forge::scikit-misc` in your active environment. Verify the environment is correctly activated.
```
For loess smoothing, install 'scikit-misc'
```
cause This error typically occurs when another library (e.g., `plotnine` or `scanpy`) tries to use `skmisc.loess` but cannot find the installed `scikit-misc` package in the execution environment.

fix Confirm `scikit-misc` is installed in the *same* Python environment where the calling library is installed and being executed. Reinstalling `scikit-misc` with `pip install --user scikit-misc --force` or ensuring the correct `PYTHONPATH` can sometimes resolve this.
```
ERROR: Command `/usr/bin/python3 -c 'import os os.chdir(os.path.join("..", "tools")) import numpy as np try: incdir = os.path.relpath(np.get_include()) except Exception: incdir = np.get_include() print(incdir) '` failed with status 1.
```
cause This is a build-time error, often related to `meson-python` failing to find `numpy` headers during compilation, particularly with newer Python versions (e.g., 3.13) where build dependencies might not be automatically linked.

fix This issue has been addressed in recent `scikit-misc` commits. Ensure you are using the latest version of `scikit-misc`. If building from source, ensure `numpy` is correctly installed in your build environment. Prefer installing pre-built wheels if available.

Warnings

gotcha Users often encounter 'ModuleNotFoundError' even after installation, particularly in complex environments like Anaconda or when using different Python interpreters.
Fix: Ensure `pip install scikit-misc` or `conda install conda-forge::scikit-misc` is run within the *correct* active Python environment. If using Jupyter, ensure the kernel points to this environment. Sometimes, reinstalling with `--user` or explicitly managing `PYTHONPATH` might be necessary.
breaking Older SciPy users might recall algorithms (e.g., `scipy.misc.face`, `scipy.misc.ascent`) that were deprecated or removed from `scipy.misc`. `scikit-misc` aims to re-introduce some of these or similar functionalities, but direct `scipy.misc` imports will no longer work and require migrating to `scikit-misc` or other SciPy submodules.
Fix: Update imports to `from skmisc.some_module import function` or use alternative functions within modern SciPy modules.
gotcha Building `scikit-misc` from source, especially on less common platforms or with pre-release Python versions (e.g., Python 3.13, or sometimes on Windows), can lead to compilation errors due to dependencies on Fortran/C/Cython code and build system interactions with `numpy`.
Fix: Use official pre-built wheels (`pip install scikit-misc`) whenever possible. For build failures, ensure `numpy` is installed as a build dependency, and check the GitHub issues for specific patches or workarounds related to your Python/Numpy version. Recent commits have addressed Python >= 3.12 and Numpy 2.0 compatibility.

Install

pip install scikit-misc Recommended via pip
conda install conda-forge::scikit-misc Recommended via Conda

Imports

loess
```
from skmisc.loess import loess
```

Quickstart

This example demonstrates how to use the `loess` function from `skmisc.loess` to perform locally-weighted regression on a sample dataset, generating smoothed predictions.

import numpy as np
from skmisc.loess import loess

# Generate some sample data
x = np.linspace(0, 10, 100)
y = np.sin(x) + np.random.normal(0, 0.5, 100)

# Fit a LOESS model
model = loess(x, y)
model.fit()

# Get smoothed predictions and statistics
pred = model.predict(x).values

print(f"First 5 original y values: {y[:5]}")
print(f"First 5 smoothed y values: {pred[:5]}")

view raw JSON →