feat: Experiment with runtime sampling

python/explore/runtime_sampling.py

@@ -0,0 +1,42 @@
+import numpy as np
+import torch
+
+
+def sample_indices(vector):
+    A = vector.copy()
+    num_ones = np.sum(A == 1)
+    print(num_ones)
+
+    sample_from = np.where(A==0)[0]
+    np.random.shuffle(sample_from)
+    picked_zeros = sample_from[0:min(num_ones, np.size(sample_from))]
+    A[picked_zeros] = 1
+
+    return np.where(A == 1)[0]
+
+
+
+
+total_num = 0.0
+count = 0
+
+for i in range(10000):
+    A = np.round(np.random.rand(30))
+    #
+    # A[np.random.randint(0,np.size(A), size=10)]=1
+
+    total_num += np.size(sample_indices(A))
+    count+=1
+
+print("Average number =", total_num/count)
+
+0/0
+
+ones = np.sum(A==1)
+
+for i in range(10000):
+    B = A.copy()
+    B[np.random.randint(0, np.size(A), size=ones)] = 2
+    print(np.sum(B==1)+np.sum(B==2))
+
+