show sample usage

README.md

@@ -20,4 +20,39 @@ Port of the compas-recidivism dataset from propublica (github [here](https://git
 
 Basic preprocessing done by the [imodels team](https://github.com/csinva/imodels) in [this notebook](https://github.com/csinva/imodels-data/blob/master/notebooks_fetch_data/00_get_datasets_custom.ipynb). 
 
-The target is the binary outcome `is_recid`.
+The target is the binary outcome `is_recid`.
+
+### Sample usage
+
+Load the data:
+
+```
+from datasets import load_dataset
+
+dataset = load_dataset("imodels/compas-recidivism")
+df = pd.DataFrame(dataset['train'])
+X = df.drop(columns=['is_recid'])
+y = df['is_recid'].values
+```
+
+Fit a model:
+
+```
+import imodels
+import numpy as np
+
+m = imodels.FIGSClassifier(max_rules=5)
+m.fit(X, y)
+print(m)
+```
+
+
+Evaluate:
+
+
+```
+df_test = pd.DataFrame(dataset['test'])
+X_test = df.drop(columns=['is_recid'])
+y_test = df['is_recid'].values
+print('accuracy', np.mean(m.predict(X_test) == y_test))
+```