X-Git-Url: https://git.njae.me.uk/?a=blobdiff_plain;f=norms.py;h=f9fc1d73aebde753a52cf5038bfa6a85db2f54ed;hb=e6332a16567643e66c2a491b94994f9482384d34;hp=4fdf1e3d85bb347c501bcb88c6caec7a8c969035;hpb=26d9d2228e47a6ff8b8696d37c0a8d6d6b906c67;p=cipher-tools.git

diff --git a/norms.py b/norms.py
index 4fdf1e3..f9fc1d7 100644
--- a/norms.py
+++ b/norms.py
@@ -13,7 +13,8 @@ def normalise(frequencies):
     [(1, 0.4082482904638631), (2, 0.8164965809277261), (3, 0.4082482904638631)]
    """
     length = sum([f ** 2 for f in frequencies.values()]) ** 0.5
-    return collections.defaultdict(int, ((k, v / length) for (k, v) in frequencies.items()))
+    return collections.defaultdict(int, ((k, v / length) 
+        for (k, v) in frequencies.items()))
 
 def scale(frequencies):
     """Scale a set of frequencies so the largest is 1
@@ -28,7 +29,8 @@ def scale(frequencies):
     [(1, 0.5), (2, 1.0), (3, 0.5)]
     """
     largest = max(frequencies.values())
-    return collections.defaultdict(int, ((k, v / largest) for (k, v) in frequencies.items()))
+    return collections.defaultdict(int, ((k, v / largest) 
+        for (k, v) in frequencies.items()))
     
 
 def l2(frequencies1, frequencies2):
@@ -97,24 +99,52 @@ def l3(frequencies1, frequencies2):
     return total ** (1/3)
 
 def geometric_mean(frequencies1, frequencies2):
-    """Finds the distances between two frequency profiles, expressed as dictionaries.
+    """Finds the geometric mean of the absolute differences between two frequency profiles, 
+    expressed as dictionaries.
     Assumes every key in frequencies1 is also in frequencies2
-
+    
+    >>> geometric_mean({'a':2, 'b':2, 'c':2}, {'a':1, 'b':1, 'c':1})
+    1
+    >>> geometric_mean({'a':2, 'b':2, 'c':2}, {'a':1, 'b':1, 'c':1})
+    1
+    >>> geometric_mean({'a':2, 'b':2, 'c':2}, {'a':1, 'b':5, 'c':1})
+    3
+    >>> geometric_mean(normalise({'a':2, 'b':2, 'c':2}), normalise({'a':1, 'b':5, 'c':1}))
+    0.057022248808851934
+    >>> geometric_mean(normalise({'a':2, 'b':2, 'c':2}), normalise({'a':1, 'b':1, 'c':1}))
+    0.0
+    >>> geometric_mean(normalise({'a':2, 'b':2, 'c':2}), normalise({'a':1, 'b':1, 'c':0}))
+    0.009720703533656434
     """
-    total = 0
+    total = 1
     for k in frequencies1.keys():
         total *= abs(frequencies1[k] - frequencies2[k])
     return total
 
 def harmonic_mean(frequencies1, frequencies2):
-    """Finds the distances between two frequency profiles, expressed as dictionaries.
+    """Finds the harmonic mean of the absolute differences between two frequency profiles, 
+    expressed as dictionaries.
     Assumes every key in frequencies1 is also in frequencies2
 
+    >>> harmonic_mean({'a':2, 'b':2, 'c':2}, {'a':1, 'b':1, 'c':1})
+    1.0
+    >>> harmonic_mean({'a':2, 'b':2, 'c':2}, {'a':1, 'b':1, 'c':1})
+    1.0
+    >>> harmonic_mean({'a':2, 'b':2, 'c':2}, {'a':1, 'b':5, 'c':1})
+    1.2857142857142858
+    >>> harmonic_mean(normalise({'a':2, 'b':2, 'c':2}), normalise({'a':1, 'b':5, 'c':1}))
+    0.3849001794597505
+    >>> harmonic_mean(normalise({'a':2, 'b':2, 'c':2}), normalise({'a':1, 'b':1, 'c':1}))
+    0
+    >>> harmonic_mean(normalise({'a':2, 'b':2, 'c':2}), normalise({'a':1, 'b':1, 'c':0}))
+    0.17497266360581604
     """
     total = 0
     for k in frequencies1.keys():
+        if abs(frequencies1[k] - frequencies2[k]) == 0:
+            return 0
         total += 1 / abs(frequencies1[k] - frequencies2[k])
-    return 1 / total
+    return len(frequencies1) / total
 
 
 def cosine_distance(frequencies1, frequencies2):