You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
🎉 MAPIE v1.0.0 is live! 🎉 You're seeing the documentation of this new version, which introduces major changes to the API. Extensive release notes are available `here <https://mapie.readthedocs.io/en/latest/v1_migration_guide.html>`_. You can switch to the documentation of previous versions using the button on the bottom right of ReadTheDoc pages.
63
42
64
-
.. code:: python
65
-
66
-
# Uncertainty quantification for classification problem
**MAPIE** is an open-source Python library for quantifying uncertainties and controlling the risks of machine learning models.
69
44
70
-
Implemented methods in **MAPIE** respect three fundamental pillars:
45
+
It allows you to:
71
46
72
-
- They are **model and use case agnostic**,
73
-
- They possess **theoretical guarantees** under minimal assumptions on the data and the model,
74
-
- They are based on **peer-reviewed algorithms** and respect programming standards.
47
+
- Easily **compute conformal prediction intervals (or prediction sets)** for regression [3,4,8], classification (binary and multi-class) [5-7],
48
+
and time series [9] with guaranteed coverage rates by using the conformity set to estimate conformity scores.
49
+
- Easily **control risks** of more complex tasks such as multi-label classification,
50
+
semantic segmentation in computer vision, providing probabilistic guarantees on metrics like recall and precision [10-12].
51
+
- Easily **integrate with any model (scikit-learn, TensorFlow, PyTorch) using a scikit-learn-compatible wrap
8000
per** if needed. MAPIE is part of the scikit-learn-contrib ecosystem.
75
52
76
-
**MAPIE** relies notably on the field of *Conformal Prediction* and *Distribution-Free Inference*.
MAPIE relies notably on the fields of Conformal Prediction and Distribution-Free Inference. It implements **peer-reviewed** algorithms that are **model and use case agnostic** and possesses **theoretical guarantees** under minimal assumptions on the data and the model.
78
58
79
-
🔗 Requirements
80
-
===============
81
59
82
-
- **MAPIE** runs on Python 3.7+.
83
-
- **MAPIE** stands on the shoulders of giants. Its only internal dependencies are `scikit-learn <https://scikit-learn.org/stable/>`_ and `numpy=>1.21 <https://numpy.org/>`_.
60
+
🛠 Requirements & installation
61
+
=================================
84
62
63
+
MAPIE runs on:
85
64
86
-
🛠 Installation
87
-
===============
65
+
- Python >=3.9, <3.12
66
+
- NumPy >=1.23
67
+
- scikit-learn >=1.4
88
68
89
-
**MAPIE** can be installed in different ways:
69
+
MAPIE can be installed in different ways:
90
70
91
71
.. code:: sh
92
72
@@ -95,86 +75,36 @@ Implemented methods in **MAPIE** respect three fundamental pillars:
95
75
$ pip install git+https://github.com/scikit-learn-contrib/MAPIE # or directly from the github repository
96
76
97
77
98
-
⚡ Quickstart
99
-
=============
100
-
101
-
Here we propose two basic uncertainty quantification problems for regression and classification tasks with scikit-learn.
102
-
103
-
As **MAPIE** is compatible with the standard scikit-learn API, you can see that with just these few lines of code:
104
-
105
-
- How easy it is **to wrap your favorite scikit-learn-compatible model** around your model.
106
-
- How easy it is **to follow the standard sequential** ``fit`` and ``predict`` process like any scikit-learn estimator.
107
-
108
-
.. code:: python
109
-
110
-
# Uncertainty quantification for regression problem
111
-
import numpy as np
112
-
from sklearn.linear_model import LinearRegression
113
-
from sklearn.datasets import make_regression
114
-
from sklearn.model_selection import train_test_split
78
+
⚡ Quickstart and documentation
79
+
=======================================
115
80
116
-
from mapie.regression import MapieRegressor
81
+
Below are two simple examples from `our documentation <https://mapie.readthedocs.io/en/latest/>`_ that show how MAPIE is used in a regression setting and a classification setting:
117
82
83
+
- `Uncertainty quantification for a regression problem <https://mapie.readthedocs.io/en/latest/examples_regression/1-quickstart/plot_toy_model.html>`_
118
84
119
-
X, y = make_regression(n_samples=500, n_features=1)
120
-
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5)
The full documentation can be found `on this link <https://mapie.readthedocs.io/en/latest/>`_.
155
-
85
+
- `Uncertainty quantification for a classification problem <https://mapie.readthedocs.io/en/latest/examples_classification/1-quickstart/plot_quickstart_classification.html>`_
156
86
157
87
📝 Contributing
158
88
===============
159
89
160
90
You are welcome to propose and contribute new ideas.
161
91
We encourage you to `open an issue <https://github.com/scikit-learn-contrib/MAPIE/issues>`_ so that we can align on the work to be done.
162
92
It is generally a good idea to have a quick discussion before opening a pull request that is potentially out-of-scope.
163
-
For more information on the contribution process, please go `here<https://github.com/scikit-learn-contrib/MAPIE/blob/master/CONTRIBUTING.rst>`_.
93
+
For more information on the contribution process, read our `contribution guidelines<https://github.com/scikit-learn-contrib/MAPIE/blob/master/CONTRIBUTING.rst>`_.
164
94
165
95
166
96
🤝 Affiliations
167
97
================
168
98
169
-
MAPIE has been developed through a collaboration between Capgemini, Quantmetry, Michelin, ENS Paris-Saclay,
99
+
MAPIE has been developed through a collaboration between Capgemini Invent, Quantmetry, Michelin, ENS Paris-Saclay,
170
100
and with the financial support from Région Ile de France and Confiance.ai.
0 commit comments