initial

utils/random_utils.py

@@ -0,0 +1,14 @@
+import numpy as np
+
+def random_normal( size=(1,), trunc_val = 2.5 ):
+    len = np.array(size).prod()
+    result = np.empty ( (len,) , dtype=np.float32)
+    
+    for i in range (len):
+        while True:
+            x = np.random.normal()
+            if x >= -trunc_val and x <= trunc_val:
+                break
+        result[i] = (x / trunc_val)
+        
+    return result.reshape ( size )