Rectified XEB R^2

Author: WrathfulSpatula

scripts/tfim_validation_qiskit.py

@@ -302,6 +302,11 @@ def main():
 
     # Add up the square residuals:
     r_squared = result["l2_difference"] ** 2
+    xeb = result["xeb"]
+    if xeb > 1.0:
+        # Rectify, to penalize values greater than 1.0
+        xeb -= xeb - 1.0
+    r_squared_xeb = (1.0 - xeb) ** 2
 
     magnetization_0, sqr_magnetization_0 = 0, 0
     for hamming_weight, value in enumerate(bias_0):
@@ -358,6 +363,11 @@ def main():
 
         # Add up the square residuals:
         r_squared += result["l2_difference"] ** 2
+        xeb = result["xeb"]
+        if xeb > 1.0:
+            # Rectify, to penalize values greater than 1.0
+            xeb -= xeb - 1.0
+        r_squared_xeb += (1.0 - xeb) ** 2
 
         magnetization, sqr_magnetization = 0, 0
         for hamming_weight, value in enumerate(bias):
@@ -384,10 +394,12 @@ def main():
     # R^2 and RMSE are elementary and standard measures of goodness-of-fit with simple definitions.
     # Ideal marginal probability would be 1.0, each depth step. Squared and summed, that's depth.
     r_squared = 1.0 - r_squared / (depth + 1)
+    r_squared_xeb = 1.0 - r_squared_xeb / (depth + 1)
     rmse = (ssr / depth) ** (1 / 2)
     sm_r_squared = 1.0 - (ssr / ss)
 
     print("L2 norm similarity R^2: " + str(r_squared))
+    print("Rectified XEB R^2: " + str(r_squared_xeb))
     print("Square magnetization RMSE: " + str(rmse))
     print("Square magnetization R^2: " + str(sm_r_squared))
 

scripts/tfim_validation_qrack.py

@@ -295,6 +295,11 @@ def main():
 
     # Add up the square residuals:
     r_squared = result["l2_difference"] ** 2
+    xeb = result["xeb"]
+    if xeb > 1.0:
+        # Rectify, to penalize values greater than 1.0
+        xeb -= xeb - 1.0
+    r_squared_xeb = (1.0 - xeb) ** 2
 
     magnetization_0, sqr_magnetization_0 = 0, 0
     for hamming_weight, value in enumerate(bias_0):
@@ -339,6 +344,11 @@ def main():
 
         # Add up the square residuals:
         r_squared += result["l2_difference"] ** 2
+        xeb = result["xeb"]
+        if xeb > 1.0:
+            # Rectify, to penalize values greater than 1.0
+            xeb -= xeb - 1.0
+        r_squared_xeb += (1.0 - xeb) ** 2
 
         magnetization, sqr_magnetization = 0, 0
         for hamming_weight, value in enumerate(bias):
@@ -365,10 +375,12 @@ def main():
     # R^2 and RMSE are elementary and standard measures of goodness-of-fit with simple definitions.
     # Ideal marginal probability would be 1.0, each depth step. Squared and summed, that's depth.
     r_squared = 1.0 - r_squared / (depth + 1)
+    r_squared_xeb = 1.0 - r_squared_xeb / (depth + 1)
     rmse = (ssr / depth) ** (1 / 2)
     sm_r_squared = 1.0 - (ssr / ss)
 
     print("L2 norm similarity R^2: " + str(r_squared))
+    print("Rectified XEB R^2: " + str(r_squared_xeb))
     print("Square magnetization RMSE: " + str(rmse))
     print("Square magnetization R^2: " + str(sm_r_squared))