classify/private/forestFindThr.cpp: return 'gain' (improvement) inste…

…ad of impurity

classify/forestTrain.m

@@ -48,7 +48,7 @@
 %
 % See also forestApply, fernsClfTrain
 %
-% Piotr's Image&Video Toolbox      Version 3.10
+% Piotr's Image&Video Toolbox      Version NEW
 % Copyright 2012 Piotr Dollar.  [pdollar-at-caltech.edu]
 % Please email me if you find bugs, or have suggestions or questions!
 % Licensed under the Simplified BSD License [see external/bsd.txt]
@@ -100,9 +100,9 @@
   % train split and continue
   fids1=wswor(fWts,F1,4); data1=data(dids1,fids1);
   [~,order1]=sort(data1); order1=uint32(order1-1);
-  [fid,thr,gini]=forestFindThr(data1,hs1,dWts(dids1),order1,H);
+  [fid,thr,gain]=forestFindThr(data1,hs1,dWts(dids1),order1,H);
   fid=fids1(fid); left=data(dids1,fid)<thr; count0=nnz(left);
-  if( gini<100 && count0>=minChild && (count(k)-count0)>=minChild )
+  if( gain>1e-10 && count0>=minChild && (count(k)-count0)>=minChild )
     child(k)=K; fids(k)=fid-1; thrs(k)=thr;
     dids{K}=dids1(left); dids{K+1}=dids1(~left);
     depth(K:K+1)=depth(k)+1; K=K+2;

classify/private/forestFindThr.cpp

@@ -12,44 +12,44 @@ inline double gini( double p ) { return p*p; }
 
 void forestFindThr( int H, int N, int F, const float *data,
   const uint32 *hs, const float *ws, const uint32 *order,
-  uint32 &fid, float &thr, double &impurity )
+  uint32 &fid, float &thr, double &gain )
 {
   int i, j, j1, j2, h; double *Wl, *Wr, *W; float *data1; uint32 *order1;
-  double impurity1, w=0, wl, wr, g=0, gl, gr;
+  double v, v0, v1, w=0, wl, wr, g=0, gl, gr;
   Wl=new double[H]; Wr=new double[H]; W=new double[H];
   for( i=0; i<H; i++ ) W[i] = 0;
   for( j=0; j<N; j++ ) { w+=ws[j]; W[hs[j]-1]+=ws[j]; }
   for( i=0; i<H; i++ ) g += gini(W[i]);
-  fid = 1; thr = 0; impurity = 1e5;
+  fid = 1; thr = 0; v0 = v = (1-g/w/w);
   for( i=0; i<F; i++ ) {
     order1=(uint32*) order+i*N; data1=(float*) data+i*size_t(N);
     for( j=0; j<H; j++ ) { Wl[j]=0; Wr[j]=W[j]; } gl=wl=0; gr=g; wr=w;
     for( j=0; j<N-1; j++ ) {
-      // gini=1-\sum_h p_h^2; impurity=ginil*wl/w+ginir*wr/w
+      // gini=1-\sum_h p_h^2; v=ginil*wl/w+ginir*wr/w
       j1=order1[j]; j2=order1[j+1]; h=hs[j1]-1;
       wl+=ws[j1]; gl-=gini(Wl[h]); Wl[h]+=ws[j1]; gl+=gini(Wl[h]);
       wr-=ws[j1]; gr-=gini(Wr[h]); Wr[h]-=ws[j1]; gr+=gini(Wr[h]);
-      impurity1 = (wl-gl/wl)/w + (wr-gr/wr)/w;
-      if( impurity1<impurity && data1[j2]-data1[j1]>=1e-6f ) {
-        impurity=impurity1; fid=i+1; thr=0.5f*(data1[j1]+data1[j2]); }
+      v1 = (wl-gl/wl)/w + (wr-gr/wr)/w;
+      if( v1<v && data1[j2]-data1[j1]>=1e-6f ) {
+        v=v1; fid=i+1; thr=0.5f*(data1[j1]+data1[j2]); }
     }
   }
-  delete [] Wl; delete [] Wr; delete [] W;
+  delete [] Wl; delete [] Wr; delete [] W; gain = v0-v;
 }
 
-// [fid,thr,impurity] = mexFunction(data,hs,ws,order,H);
+// [fid,thr,gain] = mexFunction(data,hs,ws,order,H);
 void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[]) {
   int H, N, F; float *data, *ws, thr;
-  double impurity; uint32 *hs, *order, fid;
+  double gain; uint32 *hs, *order, fid;
   data = (float*) mxGetData(prhs[0]);
   hs = (uint32*) mxGetData(prhs[1]);
   ws = (float*) mxGetData(prhs[2]);
   order = (uint32*) mxGetData(prhs[3]);
   H = (int) mxGetScalar(prhs[4]);
   N = (int) mxGetM(prhs[0]);
   F = (int) mxGetN(prhs[0]);
-  forestFindThr(H,N,F,data,hs,ws,order,fid,thr,impurity);
+  forestFindThr(H,N,F,data,hs,ws,order,fid,thr,gain);
   plhs[0] = mxCreateDoubleScalar(fid);
   plhs[1] = mxCreateDoubleScalar(thr);
-  plhs[2] = mxCreateDoubleScalar(impurity);
+  plhs[2] = mxCreateDoubleScalar(gain);
 }

external/history.txt

@@ -3,7 +3,8 @@ Version NEW
  -opticalFlow.m: various improvements, changes to parameters
  -toolboxCompile.m: added option to compile w/o openMP
  -matlab/medianw.m: added fast weighted median
- -minor: imagesAlignSeq, bbApply, convConst, imResample, gradientMex
+ -minor: imagesAlignSeq, bbApply, convConst, imResample, gradientMex,
+    forestTrain.m
 
 Version 3.23 (20-Aug-2013)
  -fhog.m: added heavily optimized FHOG (Felzenszwalb's HOG) features