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 @@
},
{
"cell_type": "code",
"execution_count": 36,
"execution_count": 6,
"metadata": {},
"outputs": [
{
......@@ -226,37 +226,37 @@
"noise 0.000000 True\n",
"Permutable 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",
"proportional increment 0.000000 True\n",
"\n",
"\n",
"--- Incremental noise on signal ---\n",
"0 0.000000 +/- 0.000000\n",
"1 0.000185 +/- 0.000440\n",
"2 0.001894 +/- 0.002282\n",
"3 0.005138 +/- 0.004323\n",
"4 0.010559 +/- 0.007702\n",
"5 0.012664 +/- 0.007561\n",
"6 0.015651 +/- 0.012095\n",
"7 0.020212 +/- 0.013827\n",
"8 0.026308 +/- 0.020254\n",
"9 0.029255 +/- 0.024522\n",
"10 0.031770 +/- 0.023523\n",
"1 0.000219 +/- 0.000476\n",
"2 0.001619 +/- 0.001852\n",
"3 0.005863 +/- 0.004867\n",
"4 0.008322 +/- 0.005766\n",
"5 0.013591 +/- 0.008541\n",
"6 0.017515 +/- 0.012170\n",
"7 0.020925 +/- 0.014721\n",
"8 0.024099 +/- 0.019659\n",
"9 0.024999 +/- 0.019011\n",
"10 0.032244 +/- 0.024834\n",
"\n",
"\n",
"--- Incremental random noise ---\n",
"0 0.000138 +/- 0.000000\n",
"1 0.025829 +/- 0.018906\n",
"2 0.033744 +/- 0.025267\n",
"3 0.028168 +/- 0.022969\n",
"4 0.028078 +/- 0.024013\n",
"5 0.029980 +/- 0.025617\n",
"6 0.028973 +/- 0.021527\n",
"7 0.028279 +/- 0.019776\n",
"8 0.028193 +/- 0.024633\n",
"9 0.029332 +/- 0.024630\n",
"10 0.025569 +/- 0.022864\n",
"1 0.025660 +/- 0.019466\n",
"2 0.027256 +/- 0.022967\n",
"3 0.030341 +/- 0.024569\n",
"4 0.027788 +/- 0.019176\n",
"5 0.023954 +/- 0.018645\n",
"6 0.025143 +/- 0.021361\n",
"7 0.030559 +/- 0.023147\n",
"8 0.025742 +/- 0.017046\n",
"9 0.026679 +/- 0.021437\n",
"10 0.026494 +/- 0.023600\n",
"\n",
"\n",
"--- Incremental unbalanced distribution ---\n",
......@@ -339,6 +339,7 @@
" am = np.argmax([pears(h,h2) for h in ps])\n",
" return am, ps[am], pears(ps[am],h2)\n",
"\n",
"\"\"\"\n",
"def probhist(hh1, hh2):\n",
" #pseudocounts\n",
" h1 = [x+1 for x in hh1]\n",
......@@ -359,6 +360,7 @@
" res += rat*math.log((rat/ratsort), 2)\n",
" return res\n",
"\n",
"\n",
"def probhist(hh1, hh2, symm=True):\n",
" #Pseudocounts\n",
" eps = 0.00001\n",
......@@ -390,33 +392,27 @@
" return res/2\n",
"\n",
" return res\n",
"\"\"\"\n",
"\n",
"\n",
"def probhist(hh1, hh2, symm=True):\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",
"def probhist(hh1, hh2, symm=True): \n",
" # Pseudocounts\n",
" eps = 0.00001\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",
" \n",
" #Normalize histos\n",
" # Normalize histos\n",
" s1, s2 = sum(h1), sum(h2)\n",
" nh1, nh2 = [x/s1 for x in h1], [x/s2 for x in h2]\n",
" \n",
" #Couple histos and sort second\n",
" # Couple histos and sort second\n",
" twonhs = list(zip(nh1,nh2))\n",
" twonhs = sorted(twonhs, key= lambda x:x[1])\n",
" \n",
" #Sort both histos and then couple\n",
" # Sort both histos and then couple\n",
" twosortednhs = list(zip(sorted(nh1), sorted(nh2)))\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",
" for i in range(len(h1)):\n",
" d = stable_sigmoid(abs(twonhs[i][0]-twonhs[i][1])-0.5)\n",
......@@ -426,8 +422,7 @@
" return res\n",
"\n",
"\n",
"\n",
"\n",
"\"\"\"\n",
"def probhist2(hh1, hh2):\n",
" def likelihood(doubleh, d):\n",
" diffs = np.zeros(len(doubleh)+1)\n",
......@@ -451,6 +446,7 @@
" 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",
" return res\n",
"\"\"\"\n",
"\n",
"def agg_probhist(h1, 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