Changes between Initial Version and Version 1 of DatabaseBasedAnalysis/RandomForest

Timestamp:

08/15/18 17:19:44 (7 years ago)

Author:

tbretz

Comment:

--

DatabaseBasedAnalysis/RandomForest

@@ +1 @@
+
+== Writing Input Files ==
+
+{{{#!Spoiler
+{{{#!cpp
+#include <iostream>
+#include <iomanip>
+#include <fstream>
+
+#include <TMath.h>
+#include <TChain.h>
+
+using namespace std;
+
+void writesim()
+{
+    // Create chain for the tree Result
+    // This is just easier than using TFile/TTree
+    TChain c("Result");
+
+    // Add the input file to the
+    c.AddFile("simulation.root");
+
+    // Define variables for all leaves to be accessed
+    // By definition rootifysql writes only doubles
+    double X, Y, MeanX, MeanY, Width, Length, CosDelta, SinDelta,
+        M3Long, SlopeLong, Leakage1, SlopeSpreadWeighted, Size,
+        ConcCore, ConcCOG, NumIslands, NumUsedPixels, Zd, Energy;
+
+    // Connect the variables to the cordesponding leaves
+    //c.SetBranchAddress("FileId", &FileId);
+    //c.SetBranchAddress("EvtNumber", &EvtNumber);
+    c.SetBranchAddress("X", &X);
+    c.SetBranchAddress("Y", &Y);
+    c.SetBranchAddress("MeanX", &MeanX);
+    c.SetBranchAddress("MeanY", &MeanY);
+    c.SetBranchAddress("Width", &Width);
+    c.SetBranchAddress("Length", &Length);
+    c.SetBranchAddress("CosDelta", &CosDelta);
+    c.SetBranchAddress("SinDelta", &SinDelta);
+    c.SetBranchAddress("M3Long", &M3Long);
+    c.SetBranchAddress("SlopeLong", &SlopeLong);
+    c.SetBranchAddress("Leakage1", &Leakage1);
+    c.SetBranchAddress("NumIslands", &NumIslands);
+    c.SetBranchAddress("NumUsedPixels", &NumUsedPixels);
+    c.SetBranchAddress("Size", &Size);
+    c.SetBranchAddress("Zd", &Zd);
+    c.SetBranchAddress("Energy", &Energy);
+
+    // Set some constants (they could be included in the database
+    // in the future)
+    double mm2deg = +0.0117193246260285378;
+    //double abberation = 1.02;
+
+    // -------------------- Source dependent parameter calculation -------------------
+
+    ofstream fout0("sim-train.csv"); // %1
+    ofstream fout1("sim-test.csv");  // %0
+    ofstream fout2("sim-test-cuts.csv");
+
+    fout0 << "Energy Size Zd Dist Disp Slope M3L Leakage Width Length" << endl;
+    fout1 << "Energy Size Zd Dist Disp Slope M3L Leakage Width Length" << endl;
+    fout2 << "Energy Size Zd Dist Disp Slope M3L Leakage Width Length" << endl;
+
+    // Loop over all wobble positions in the camera
+    for (int i=0; i<c.GetEntries(); i++)
+    {
+        // read the i-th event from the file
+        c.GetEntry(i);
+
+        // First calculate all cuts to speedup the analysis
+        double area = TMath::Pi()*Width*Length;
+
+        // The abberation correction does increase also Width and Length by 1.02
+
+        int angle = 0;
+
+        // -------------------- Source dependent parameter calculation -------------------
+
+        double cr = cos(angle*TMath::DegToRad());
+        double sr = sin(angle*TMath::DegToRad());
+
+        double px = cr*X-sr*Y;
+        double py = cr*Y+sr*X;
+
+        double dx = MeanX - px*1.022;
+        double dy = MeanY - py*1.022;
+
+        double norm = sqrt(dx*dx + dy*dy);
+        double dist = norm*mm2deg;
+
+        double lx = min(max((CosDelta*dy - SinDelta*dx)/norm, -1.), 1.);
+        double ly = min(max((CosDelta*dx + SinDelta*dy)/norm, -1.), 1.);
+
+        double alpha = asin(lx);
+        double sgn   = TMath::Sign(1., ly);
+
+        // ------------------------------- Application ----------------------------------
+
+        double m3l   = M3Long*sgn*mm2deg;
+        double slope = SlopeLong*sgn/mm2deg;
+
+        // --------------------------------- Analysis -----------------------------------
+
+        //double xi = 1.34723 + 0.15214 *slope + 0.970704*(1-1/(1+8.89826*Leakage1));
+        double xi = 1.340 + 0.0755*slope + 1.67972*(1-1/(1+4.86232*Leakage1));
+
+        double sign1 = m3l+0.07;
+        double sign2 = (dist-0.5)*7.2-slope;
+
+        double disp  = (sign1<0 || sign2<0 ? -xi : xi)*(1-Width/Length);
+
+        double thetasq = disp*disp + dist*dist - 2*disp*dist*sqrt(1-lx*lx);
+
+        if (i%2==0)
+        {
+            fout0 << log10(Energy) << " ";
+            fout0 << log10(Size) << " ";
+            fout0 << Zd << " ";
+            fout0 << dist << " ";
+            fout0 << disp << " ";
+            fout0 << slope << " ";
+            fout0 << m3l << " ";
+            fout0 << Leakage1 << " ";
+            fout0 << Width << " ";
+            fout0 << Length << endl;
+        }
+        else
+        {
+            fout1 << log10(Energy) << " ";
+            fout1 << log10(Size) << " ";
+            fout1 << Zd << " ";
+            fout1 << dist << " ";
+            fout1 << disp << " ";
+            fout1 << slope << " ";
+            fout1 << m3l << " ";
+            fout1 << Leakage1 << " ";
+            fout1 << Width << " ";
+            fout1 << Length << endl;
+
+            if (thetasq<0.024)
+                continue;
+
+            bool cutq = NumIslands<3.5 && NumUsedPixels>5.5 && Leakage1<0.1;
+            if (!cutq)
+                continue;
+
+            bool cut0 = area < log10(Size)*898-1535;
+            if (!cut0)
+                continue;
+
+            fout2 << log10(Energy) << " ";
+            fout2 << log10(Size) << " ";
+            fout2 << Zd << " ";
+            fout2 << dist << " ";
+            fout2 << disp << " ";
+            fout2 << slope << " ";
+            fout2 << m3l << " ";
+            fout2 << Leakage1 << " ";
+            fout2 << Width << " ";
+            fout2 << Length << endl;
+        }
+    }
+}
+}}}
+}}}
+
+== Training ==
+
+{{{
+fact@ihp-pc45:~/Analysis> nice -n 10 ~/ranger-master/cpp_version/build/ranger --file sim-train.csv --depvarname Energy --memmode 1 --treetype 3 --verbose --impmeasure 1 --outprefix sim-train
+Starting Ranger.
+Loading input file: sim-train.csv.
+Growing trees ..
+Computing prediction error ..
+
+Tree type:                         Regression
+Dependent variable name:           Energy
+Dependent variable ID:             0
+Number of trees:                   500
+Sample size:                       55417
+Number of independent variables:   9
+Mtry:                              3
+Target node size:                  5
+Variable importance mode:          1
+Memory mode:                       1
+Seed:                              0
+Number of threads:                 8
+
+Overall OOB prediction error:      0.0178514
+
+Saved variable importance to file sim-train.importance.
+Saved prediction error to file sim-train.confusion.
+Finished Ranger.
+}}}
+
+It will write a file called '''sim-train.forest'''.
+
+== Testing ==
+
+{{{
+nice -n 10 ~/ranger-master/cpp_version/build/ranger --file sim-test.csv      --depvarname Energy --memmode 1 --treetype 3 --verbose --impmeasure 1 --predict sim-train.forest 
+nice -n 10 ~/ranger-master/cpp_version/build/ranger --file sim-test-cuts.csv --depvarname Energy --memmode 1 --treetype 3 --verbose --impmeasure 1 --predict sim-train.forest 
+}}}
+
+Here is an example output
+{{{
+Starting Ranger.
+Loading input file: sim-test-cuts.csv.
+Loading forest from file sim-train.forest.
+Predicting ..
+
+Tree type:                         Regression
+Dependent variable name:           Energy
+Dependent variable ID:             0
+Number of trees:                   500
+Sample size:                       5135
+Number of independent variables:   9
+Mtry:                              3
+Target node size:                  5
+Variable importance mode:          1
+Memory mode:                       1
+Seed:                              0
+Number of threads:                 8
+
+Saved predictions to file ranger_out.prediction.
+Finished Ranger.
+}}}
+
+
+