Began work on plotting improvements between training runs

plot_improvement.py

@@ -0,0 +1,42 @@
+import argparse
+from numpy import load, ndarray
+
+parser = argparse.ArgumentParser()
+parser.add_argument("-f", "--filepath", required=True, help="Specify the file path to the agent.", type=str)
+args = parser.parse_args()
+
+filepath = args.filepath
+npdata = load(filepath)
+
+print(type(npdata['results']))
+evaluations = ndarray.tolist(npdata['results'])
+print(type(evaluations))
+print(len(evaluations))
+# print(evaluations)
+sorted_evals = []
+for eval in evaluations:
+    sorted_evals.append(sorted(eval))
+
+# Now I have a sorted list.
+# Now just pop the first and last elements of each eval
+for eval in sorted_evals:
+    eval.pop(0)
+    eval.pop()
+
+print()
+# print(sorted_evals)
+
+# Now that I have my sorted evaluations, I can calculate the mean episode reward for each eval
+mean_eval_rewards = []
+for eval in sorted_evals:
+    mean_eval_rewards.append(sum(eval) / len(eval))
+    
+# Now I should have a list with the mean evaluation reward with the highest and lowest score tossed out
+print(mean_eval_rewards)
+print("num evals: " + str(len(mean_eval_rewards)))
+
+# I'm dealing with a 2D array. Each element contains an array of ten data points
+# The number of elements is going to vary for each training run
+# The number of evaluation episodes will be constant, 10.
+# I need to convert to a regular list first
+# I could iterate over each element