Skip to content

P
profileview
Commit c3e61ba4 by Edoardo Sarti
Options

cleanup

parent 27a3d69a
Showing with 34 additions and 38 deletions
tree_similarity/mesureHistoramme.ipynb
...@@ -212,7 +212,7 @@ ...@@ -212,7 +212,7 @@
}, },
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 36, "execution_count": 6,
"metadata": {}, "metadata": {},
"outputs": [ "outputs": [
{ {
...@@ -226,37 +226,37 @@ ...@@ -226,37 +226,37 @@
"noise 0.000000 True\n", "noise 0.000000 True\n",
"Permutable 0.000000 True\n", "Permutable 0.000000 True\n",
"non-proportional increment 0.000000 True\n", "non-proportional increment 0.000000 True\n",
"Random big numbers or zero 0.025459 False\n", "Random big numbers or zero 0.024565 False\n",
"max unbalance 0.146087 False\n", "max unbalance 0.146087 False\n",
"proportional increment 0.000000 True\n", "proportional increment 0.000000 True\n",
"\n", "\n",
"\n", "\n",
"--- Incremental noise on signal ---\n", "--- Incremental noise on signal ---\n",
"0 0.000000 +/- 0.000000\n", "0 0.000000 +/- 0.000000\n",
"1 0.000185 +/- 0.000440\n", "1 0.000219 +/- 0.000476\n",
"2 0.001894 +/- 0.002282\n", "2 0.001619 +/- 0.001852\n",
"3 0.005138 +/- 0.004323\n", "3 0.005863 +/- 0.004867\n",
"4 0.010559 +/- 0.007702\n", "4 0.008322 +/- 0.005766\n",
"5 0.012664 +/- 0.007561\n", "5 0.013591 +/- 0.008541\n",
"6 0.015651 +/- 0.012095\n", "6 0.017515 +/- 0.012170\n",
"7 0.020212 +/- 0.013827\n", "7 0.020925 +/- 0.014721\n",
"8 0.026308 +/- 0.020254\n", "8 0.024099 +/- 0.019659\n",
"9 0.029255 +/- 0.024522\n", "9 0.024999 +/- 0.019011\n",
"10 0.031770 +/- 0.023523\n", "10 0.032244 +/- 0.024834\n",
"\n", "\n",
"\n", "\n",
"--- Incremental random noise ---\n", "--- Incremental random noise ---\n",
"0 0.000138 +/- 0.000000\n", "0 0.000138 +/- 0.000000\n",
"1 0.025829 +/- 0.018906\n", "1 0.025660 +/- 0.019466\n",
"2 0.033744 +/- 0.025267\n", "2 0.027256 +/- 0.022967\n",
"3 0.028168 +/- 0.022969\n", "3 0.030341 +/- 0.024569\n",
"4 0.028078 +/- 0.024013\n", "4 0.027788 +/- 0.019176\n",
"5 0.029980 +/- 0.025617\n", "5 0.023954 +/- 0.018645\n",
"6 0.028973 +/- 0.021527\n", "6 0.025143 +/- 0.021361\n",
"7 0.028279 +/- 0.019776\n", "7 0.030559 +/- 0.023147\n",
"8 0.028193 +/- 0.024633\n", "8 0.025742 +/- 0.017046\n",
"9 0.029332 +/- 0.024630\n", "9 0.026679 +/- 0.021437\n",
"10 0.025569 +/- 0.022864\n", "10 0.026494 +/- 0.023600\n",
"\n", "\n",
"\n", "\n",
"--- Incremental unbalanced distribution ---\n", "--- Incremental unbalanced distribution ---\n",
...@@ -339,6 +339,7 @@ ...@@ -339,6 +339,7 @@
" am = np.argmax([pears(h,h2) for h in ps])\n", " am = np.argmax([pears(h,h2) for h in ps])\n",
" return am, ps[am], pears(ps[am],h2)\n", " return am, ps[am], pears(ps[am],h2)\n",
"\n", "\n",
"\"\"\"\n",
"def probhist(hh1, hh2):\n", "def probhist(hh1, hh2):\n",
" #pseudocounts\n", " #pseudocounts\n",
" h1 = [x+1 for x in hh1]\n", " h1 = [x+1 for x in hh1]\n",
...@@ -359,6 +360,7 @@ ...@@ -359,6 +360,7 @@
" res += rat*math.log((rat/ratsort), 2)\n", " res += rat*math.log((rat/ratsort), 2)\n",
" return res\n", " return res\n",
"\n", "\n",
"\n",
"def probhist(hh1, hh2, symm=True):\n", "def probhist(hh1, hh2, symm=True):\n",
" #Pseudocounts\n", " #Pseudocounts\n",
" eps = 0.00001\n", " eps = 0.00001\n",
...@@ -390,33 +392,27 @@ ...@@ -390,33 +392,27 @@
" return res/2\n", " return res/2\n",
"\n", "\n",
" return res\n", " return res\n",
"\"\"\"\n",
"\n", "\n",
"\n", "def probhist(hh1, hh2, symm=True): \n",
"def probhist(hh1, hh2, symm=True):\n", " # Pseudocounts\n",
" def likelihood(doubleh, d):\n",
" diffs = np.zeros(len(doubleh)+1)\n",
" for i in range(len(doubleh)):\n",
" diffs[i] = abs(doubleh[i][0]-doubleh[i][1])\n",
" diffs[-1] = 10\n",
" return sum(diffs >= d)/(len(diffs))\n",
" \n",
" #NO Pseudocounts\n",
" eps = 0.00001\n", " eps = 0.00001\n",
" h1 = [x+(max(hh1)-min(hh1)+eps)/100 for x in hh1]\n", " h1 = [x+(max(hh1)-min(hh1)+eps)/100 for x in hh1]\n",
" h2 = [x+(max(hh2)-min(hh2)+eps)/100 for x in hh2]\n", " h2 = [x+(max(hh2)-min(hh2)+eps)/100 for x in hh2]\n",
" \n", " \n",
" #Normalize histos\n", " # Normalize histos\n",
" s1, s2 = sum(h1), sum(h2)\n", " s1, s2 = sum(h1), sum(h2)\n",
" nh1, nh2 = [x/s1 for x in h1], [x/s2 for x in h2]\n", " nh1, nh2 = [x/s1 for x in h1], [x/s2 for x in h2]\n",
" \n", " \n",
" #Couple histos and sort second\n", " # Couple histos and sort second\n",
" twonhs = list(zip(nh1,nh2))\n", " twonhs = list(zip(nh1,nh2))\n",
" twonhs = sorted(twonhs, key= lambda x:x[1])\n", " twonhs = sorted(twonhs, key= lambda x:x[1])\n",
" \n", " \n",
" #Sort both histos and then couple\n", " # Sort both histos and then couple\n",
" twosortednhs = list(zip(sorted(nh1), sorted(nh2)))\n", " twosortednhs = list(zip(sorted(nh1), sorted(nh2)))\n",
" \n", " \n",
" # SUM P1/P2 log2(P1/P1') = SUM P1/P2 log2((P1/P2) / (P1'/P2')) because P2 = P2'\n", " # Cross-mutual information\n",
" # SUM f(h1,h2) log2 (f(h1,h2) / f(h1',h2')) with h2 = h2'\n",
" res = 0\n", " res = 0\n",
" for i in range(len(h1)):\n", " for i in range(len(h1)):\n",
" d = stable_sigmoid(abs(twonhs[i][0]-twonhs[i][1])-0.5)\n", " d = stable_sigmoid(abs(twonhs[i][0]-twonhs[i][1])-0.5)\n",
...@@ -426,8 +422,7 @@ ...@@ -426,8 +422,7 @@
" return res\n", " return res\n",
"\n", "\n",
"\n", "\n",
"\n", "\"\"\"\n",
"\n",
"def probhist2(hh1, hh2):\n", "def probhist2(hh1, hh2):\n",
" def likelihood(doubleh, d):\n", " def likelihood(doubleh, d):\n",
" diffs = np.zeros(len(doubleh)+1)\n", " diffs = np.zeros(len(doubleh)+1)\n",
...@@ -451,6 +446,7 @@ ...@@ -451,6 +446,7 @@
" for i in range(len(h1)):\n", " for i in range(len(h1)):\n",
" res += likelihood(twosortednhs, twonhs[i][0]-twonhs[i][1])*math.log(likelihood(twosortednhs, twonhs[i][0]-twonhs[i][1]), 2)\n", " res += likelihood(twosortednhs, twonhs[i][0]-twonhs[i][1])*math.log(likelihood(twosortednhs, twonhs[i][0]-twonhs[i][1]), 2)\n",
" return res\n", " return res\n",
"\"\"\"\n",
"\n", "\n",
"def agg_probhist(h1, h2):\n", "def agg_probhist(h1, h2):\n",
" return probhist(agg_vector(h1), agg_vector(h2))\n", " return probhist(agg_vector(h1), agg_vector(h2))\n",
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment