Restore scorer plotting and fix DataFrame conversion

reachml/scoring.py

@@ -2,15 +2,22 @@
 
 from abc import ABC, abstractmethod
 from collections import defaultdict
+from typing import List
 
+import matplotlib.pyplot as plt
 import numpy as np
-from typing import List
+import pandas as pd
 from scipy.sparse import csr_matrix
 from tqdm import tqdm
 
 from .action_set import ActionSet
 from .mip import EnumeratorMIP
 
+RESP_BAR_COLOR = "#FFC000"
+
+# matplotlib font params
+plt.rcParams["font.size"] = 15
+
 
 class ResponsivenessScorer(ABC):
     """Base scorer that computes per-feature responsiveness for a point."""
@@ -76,6 +83,30 @@ def _get_inter_key(self, x):
         """Return a hashable key for caching interventions for `x`."""
         pass
 
+    def plot(self, score_lst=None, x_idx=None):
+        """Plot responsiveness scores given a point."""
+        if score_lst is None and x_idx is None:
+            raise ValueError("Either scores or x_idx must be provided")
+
+        if score_lst is None:
+            score_lst = self.scores[x_idx]
+
+        # Sort the scores and names together
+        names = self.action_set.names
+        sorted_data = sorted(
+            zip(score_lst, names, strict=True), reverse=False
+        )  # Sort by score descending
+        sorted_scores, sorted_names = zip(*sorted_data, strict=True)
+
+        fig, ax = plt.subplots(figsize=(8, 6))
+
+        ax.barh(sorted_names, sorted_scores, color=RESP_BAR_COLOR)
+        ax.set_xlabel("Responsiveness Score")
+        ax.set_yticks(range(len(sorted_scores)))
+        ax.set_xlim(0, 1)
+
+        return fig
+
     def __call__(self, X, clf, save=True, **kwargs):
         """Compute per-feature responsiveness scores for each row in `X`.
 
@@ -88,6 +119,9 @@ def __call__(self, X, clf, save=True, **kwargs):
         Returns:
             Numpy array of shape (n_samples, n_features) with responsiveness scores.
         """
+        if isinstance(X, pd.DataFrame):
+            X = X.to_numpy()
+
         out = np.zeros((len(X), len(self.action_set)))
 
         for i, x in tqdm(enumerate(X), total=len(X)):