Installation#
Empulse requires Python 3.11 or higher.
pip install empulse
That gives you every metric, the linear and tree-based models, the samplers and the optimizers.
Optional extras#
Two features need extra dependencies, both installable as extras:
Extra |
Install |
What it unlocks |
|---|---|---|
|
|
XGBoost, LightGBM and CatBoost, the backends behind
|
|
|
gplearn, required by |
To install everything:
pip install empulse[optional,symbolic]
If you use a backend that is not installed, Empulse raises an error telling you exactly what to install — nothing fails silently.
Dataframe support#
The bundled datasets in empulse.datasets return dataframes, so they need pandas or
polars installed. Neither is a hard dependency of Empulse itself, because the rest of the package
works fine on plain NumPy arrays.
pip install pandas
Every loader takes a required, keyword-only backend argument — you pass the library module
itself, which is how Empulse stays agnostic between pandas and polars:
import pandas as pd
from empulse.datasets import fetch_iranian_churn
dataset = fetch_iranian_churn(backend=pd)
Verifying the install#
import empulse
print(empulse.__version__)
Note that Empulse has no top-level re-exports: import from the submodules
(empulse.metrics, empulse.models, empulse.samplers,
empulse.optimizers, empulse.datasets) rather than from empulse directly.
Next steps#
Quickstart — a working cost-sensitive model in five minutes.
Which tool do I need? — which part of Empulse solves your problem.