1 files changed, 53 insertions, 0 deletions

diff --git a/Metrics/Metrics-Calculation/metrics_plot/utils/DistributionMetrics.py b/Metrics/Metrics-Calculation/metrics_plot/utils/DistributionMetrics.py
new file mode 100644
index 00000000..cf532bc5
--- /dev/null
+++ b/Metrics/Metrics-Calculation/metrics_plot/utils/DistributionMetrics.py
@@ -0,0 +1,53 @@
+from scipy import stats
+from scipy.spatial import distance
+
+def ks_distance(samples1, samples2):
+    value, p = stats.ks_2samp(samples1, samples2)
+    return (value, p)
+
+def manual_ks(pdf1, pdf2):
+    result = 0
+    sum1 = 0
+    sum2 = 0
+    for(a, b) in zip(pdf1, pdf2):
+        sum1 += a
+        sum2 += b
+        result = max(result, abs(sum1-sum2))
+    return result
+
+def js_distance(samples1, samples2):
+    map1 = fromSamples(samples1)
+    map2 = fromSamples(samples2)
+    allKeys = set(map1.keys()) | set(map2.keys())
+    dist1 = distributionFromMap(map1, allKeys)
+    dist2 = distributionFromMap(map2, allKeys)
+    return distance.jensenshannon(dist1, dist2, 2)
+
+def euclidean_distance(samples1, samples2):
+    map1 = fromSamples(samples1)
+    map2 = fromSamples(samples2)
+    allKeys = set(map1.keys()) | set(map2.keys())
+    dist1 = distributionFromMap(map1, allKeys)
+    dist2 = distributionFromMap(map2, allKeys)
+    distance = 0
+    for i in range(len(dist2)):
+        distance += pow(dist1[i] - dist2[i], 2)
+    return pow(distance, 0.5)
+
+def fromSamples(samples):
+    m = {}
+    length = len(samples)
+    for sample in samples:
+        value = m.get(sample, 0)
+        m[sample] = value + 1
+    for key in list(m.keys()):
+        m[key] /= length
+    return m
+
+def distributionFromMap(m, allKeys):
+    dist = []
+    for key in allKeys:
+        value = m.get(key, 0)
+        dist.append(value)
+    return dist
+