Merge https://github.com/clstaudt/graphchi-cpp

Author: clstaudt

README.md

@@ -2,6 +2,8 @@
 
 **NOTE: This project has been recently moved from Google Code, and some of the wiki pages might be partly broken.**
 
+MIT Technology Review article about GraphChi: ["Your laptop can now analyze big data"](http://www.technologyreview.com/news/428497/your-laptop-can-now-analyze-big-data/?nlid=nldly&nld=2012-07-17)
+
 ## NEW: GraphChi version 0.2
 
 Highlight of the release includes support for dynamically sized edge and vertex values.

toolkits/collaborative_filtering/common.hpp

@@ -38,7 +38,7 @@
 
 #include <stdio.h>
 #ifdef __APPLE__
-#include "getline.hpp" //fix for missing getline() function on MAC OS
+//#include "getline.hpp" //fix for missing getline() function on MAC OS
 #endif 
 
 using namespace graphchi;

toolkits/collaborative_filtering/gensgd.cpp

@@ -40,17 +40,17 @@
 #include "../parsers/common.hpp"
 #include <omp.h>
 #define MAX_FEATURES 256
-#define FEATURE_WIDTH 11 //MAX NUMBER OF ALLOWED FEATURES IN TEXT FILE
-
-double gensgd_rate1 = 1e-02;
-double gensgd_rate2 = 1e-02;
-double gensgd_rate3 = 1e-02;
-double gensgd_rate4 = 1e-02;
-double gensgd_rate5 = 1e-02;
-double gensgd_mult_dec = 0.9;
+#define FEATURE_WIDTH 24 //MAX NUMBER OF ALLOWED FEATURES IN TEXT FILE
+
+double gensgd_rate1 = 1e-03;
+double gensgd_rate2 = 1e-03;
+double gensgd_rate3 = 1e-03;
+double gensgd_rate4 = 1e-03;
+double gensgd_rate5 = 1e-03;
+double gensgd_mult_dec = 0.9999999;
 double gensgd_regw = 1e-3;
 double gensgd_regv = 1e-3;
-double gensgd_reg0 = 1e-1;
+double gensgd_reg0 = 1e-2;
 bool debug = false;
 std::string user_file; //optional file with user features
 std::string item_file; //optional file with item features
@@ -1290,6 +1290,8 @@ int main(int argc, const char ** argv) {
       logstream(LOG_FATAL)<<"Feature id using the --features=XX command should be non negative, starting from zero"<<std::endl;
     if (node >= file_columns)
       logstream(LOG_FATAL)<<"Feature id using the --feature=XX command should be < file_columns (counting starts from zero)" << std::endl;
+    if (node == fc.from_pos || node == fc.to_pos || node == fc.val_pos)
+      logstream(LOG_FATAL)<<"Feature id " << node << " can not be equal to --from_pos, --to_pos or --val_pos " << std::endl;
     fc.feature_selection[node] = true;
     fc.total_features++;
     while ((pch = strtok(NULL, ",\n\r\t "))!= NULL){

toolkits/collaborative_filtering/pmf.cpp

@@ -372,7 +372,7 @@ struct PMFVerticesInMemProgram : public GraphChiProgram<VertexDataType, EdgeData
     if (iteration >= pmf_burn_in){
       rmse_index = 0;
       rmse_type = TEST;
-      test_predictions(&pmf_predict, &gcontext, iiter == niters-1, &test_avgprod);
+      test_predictions(&pmf_predict, &gcontext, iiter == niters-1, &test_avgprod, pmf_burn_in);
     }
     iiter++;
   }

toolkits/collaborative_filtering/rmse.hpp

@@ -98,7 +98,7 @@ void test_predictions(float (*prediction_func)(const vertex_data & user, const v
   size_t nz;   
 
   if ((f = fopen(test.c_str(), "r")) == NULL) {
-    return; //missing validaiton data, nothing to compute
+    return; //missing test data, nothing to compute
   }
   FILE * fout = NULL;
   if (dosave)