From 59091fda4f141b52d3655cab00f3a22b5d8d23cf Mon Sep 17 00:00:00 2001
From: baniasbaabe <banias@hotmail.de>
Date: Mon, 15 Jan 2024 18:33:14 +0100
Subject: [PATCH] Add mapie

---
 book/machinelearning/modeltraining.ipynb | 52 ++++++++++++++++++++++++
 1 file changed, 52 insertions(+)

diff --git a/book/machinelearning/modeltraining.ipynb b/book/machinelearning/modeltraining.ipynb
index c0353dd..d8bcee6 100644
--- a/book/machinelearning/modeltraining.ipynb
+++ b/book/machinelearning/modeltraining.ipynb
@@ -1837,6 +1837,58 @@
     "model = timm.create_model('densenet121', pretrained=True)\n",
     "output = model(torch.randn(2, 3, 224, 224))"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "69feb793",
+   "metadata": {},
+   "source": [
+    "## Generate Guaranteed Prediction Intervals and Sets with `MAPIE`"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "41226b58",
+   "metadata": {},
+   "source": [
+    "For quantifying uncertainties of your models, use MAPIE.\n",
+    "\n",
+    "`MAPIE` (Model Agnostic Prediction Interval Estimator) takes your sklearn-/tensorflow-/pytorch-compatible model and generate prediction intervals or sets with guaranteed coverage."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "f7abfe51",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!pip install mapie"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "ecf3cdf1",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from mapie.regression import MapieRegressor\n",
+    "import numpy as np\n",
+    "from sklearn.linear_model import LinearRegression\n",
+    "from sklearn.datasets import make_regression\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "\n",
+    "X, y = make_regression(n_samples=500, n_features=1, noise=20, random_state=59)\n",
+    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5)\n",
+    "regressor = LinearRegression()\n",
+    "\n",
+    "mapie_regressor = MapieRegressor(regressor)\n",
+    "mapie_regressor.fit(X_train, y_train)\n",
+    "\n",
+    "alpha = [0.05, 0.20]\n",
+    "y_pred, y_pis = mapie_regressor.predict(X_test, alpha=alpha)"
+   ]
   }
  ],
  "metadata": {