epsilon-compositionality

commit a02677026c33d217b07506ec99bfea99f20c1d63
parent 9212ccc92b52c0b9d06181648ebc8e5a878eb8a8
Author: Sondre Wold <sondrewold1@gmail.com>
Date:   Tue, 26 Aug 2025 09:00:27 +0200

save distribution to json instead of csv

Diffstat:
M
get_epsilon.py
 | 
47
+++++++++++++++++++++++++++++------------------

1 file changed, 29 insertions(+), 18 deletions(-)
diff --git a/get_epsilon.py b/get_epsilon.py
@@ -6,6 +6,7 @@ from experiment import Point
 import matplotlib.pyplot as plt
 
 from typing import Any
+import json
 
 import pandas
 
@@ -21,6 +22,7 @@ def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(description="Gather epsilon values")
     parser.add_argument("-points_path", type=pathlib.Path)
     parser.add_argument("-result_path", type=pathlib.Path)
+    parser.add_argument("-percentile", type=float, default=None, help="The percentile for the epsilon value")
     return parser.parse_args()
 
 
@@ -32,25 +34,22 @@ def get_dataset_and_model_name(filename: str) -> tuple[str, str]:
 
 
 def get_epsilons(
-    word_sims: numpy.ndarray, context_sims: numpy.ndarray, percentile: int = 80
-) -> dict[str, float]:
+    word_sims: numpy.ndarray, context_sims: numpy.ndarray
+) -> dict[str, list[float]]:
     # Epsilons
     original: list[float] = []
     division: list[float] = []
     additive: list[float] = []
     for i in range(len(word_sims)):
         if word_sims[i] > 0:
-            original.append(1 - context_sims[i] / word_sims[i])  # Equivalent to our formal relaxation
-            division.append(context_sims[i] / word_sims[i])
-            additive.append(context_sims[i] - word_sims[i])
-    final_original: float = float(numpy.percentile(original, percentile))
-    final_division: float = float(numpy.percentile(division, percentile))
-    final_additive: float = float(numpy.percentile(additive, percentile))
+            original.append(float(1 - context_sims[i] / word_sims[i]))  # Equivalent to our formal relaxation)
+            division.append(float(context_sims[i] / word_sims[i]))
+            additive.append(float(context_sims[i] - word_sims[i]))
 
     return {
-        "original": final_original,
-        "divisive": final_division,
-        "additive": final_additive,
+        "original": original,
+        "divisive": division,
+        "additive": additive,
     }
 
 
@@ -59,6 +58,7 @@ def main() -> None:
     points_folder: str = args.points_path
     raw_files = list(pathlib.Path(points_folder).glob("*.pkl"))
 
+
     final_result: dict[str, Any] = {
         "dataset": [],
         "model": [],
@@ -77,15 +77,26 @@ def main() -> None:
             context_sims: numpy.ndarray = numpy.array(
                 [point.contextualized_similarity for point in points]
             )
-            tmp_result: dict[str, float] = get_epsilons(word_sims, context_sims)
+            tmp_result: dict[str, list[float]] = get_epsilons(word_sims, context_sims)
             final_result["dataset"].append(dataset_name)
             final_result["model"].append(model_name)
-            final_result["original"].append(tmp_result["original"])
-            final_result["divisive"].append(tmp_result["divisive"])
-            final_result["additive"].append(tmp_result["additive"])
-
-    df: pandas.DataFrame = pandas.DataFrame.from_dict(final_result)
-    df.to_csv(args.result_path, sep=",", index=False)
+            if args.percentile is not None:  # Save the percentile epsilons to disk as a csv
+                final_original: Any = float(numpy.percentile(tmp_result["original"], args.percentile))
+                final_division: Any = float(numpy.percentile(tmp_result["divisive"], args.percentile))
+                final_additive: Any = float(numpy.percentile(tmp_result["additive"], args.percentile))
+            else:
+                final_original = tmp_result["original"]
+                final_division = tmp_result["divisive"]
+                final_additive = tmp_result["additive"]
+
+            final_result["original"].append(final_original)
+            final_result["divisive"].append(final_division)
+            final_result["additive"].append(final_additive)
+
+    #df: pandas.DataFrame = pandas.DataFrame.from_dict(final_result)
+    #df.to_csv(args.result_path, sep=",", index=False)
+    with open(args.result_path, "w") as f:
+        json.dump(final_result, f)
 
 
 if __name__ == "__main__":