/**
* Grab pivot's adjacency list into memory.
*/
int load_edges_into_memory(graphchi_vertex<uint32_t, edge_data> &v) {
assert(is_pivot(v.id()));
assert(is_user(v.id()));
int num_edges = v.num_edges();
dense_adj dadj;
for(int i=0; i<num_edges; i++)
set_new( dadj.edges, v.edge(i)->vertex_id(), v.edge(i)->get_data().up_weight);
//dadj.ratings = zeros(N);
dadj.vid = v.id();
adjs[v.id() - pivot_st] = dadj;
assert(v.id() - pivot_st < adjs.size());
__sync_add_and_fetch(&grabbed_edges, num_edges /*edges_to_larger_id*/);
return num_edges;
}
/**
* add weighted ratings for each linked item
*
*/
double compute_ratings(graphchi_vertex<uint32_t, edge_data> &item, vid_t user_pivot, float user_item_edge_weight) {
assert(is_pivot(user_pivot));
if (!allow_zeros)
assert(user_item_edge_weight != 0);
else {
if (user_item_edge_weight == 0)
return 0;
}
dense_adj &pivot_edges = adjs[user_pivot - pivot_st];
if (!get_val(pivot_edges.edges, item.id())){
if (debug)
std::cout<<"Skipping item pivot pair since not connected!" << item.id() << std::endl;
return 0;
}
int num_edges = item.num_edges();
if (debug)
std::cout<<"Found " << num_edges << " edges from item : " << item.id()-M+1 << std::endl;
//if there are not enough neighboring user nodes to those two items there is no need
//to actually count the intersection
if (num_edges < min_allowed_intersection || nnz(pivot_edges.edges) < min_allowed_intersection){
if (debug)
std::cout<<"skipping item pivot pair since < min_allowed_intersection" << std::endl;
return 0;
}
for(int i=0; i < num_edges; i++){
vid_t other_item = item.edge(i)->vertex_id();
//user node, continue
if (other_item < M ){
if (debug)
std::cout<<"skipping an edge to item " << other_item << std::endl;
continue;
}
//other item node
assert(is_item(other_item));
assert(other_item != item.id());
bool up = item.id() < other_item;
if (debug)
std::cout<<"Checking now edge number " << i << " " << item.id()-M+1 << " -> " << other_item-M+1 << " weight: " <<
item.edge(i)->get_data().up_weight + item.edge(i)->get_data().down_weight << std::endl;
if ((up && item.edge(i)->get_data().up_weight == 0) ||
(!up && item.edge(i)->get_data().down_weight == 0)){
if (debug)
std::cout<<"skipping edge with wrong direction to " << other_item-M+1 << std::endl;
continue;
}
if (get_val(pivot_edges.edges, other_item)){
if (debug)
std::cout<<"skipping edge to " << other_item << " because alrteady connected to pivot" << std::endl;
continue;
}
assert(get_val(pivot_edges.edges, item.id()) != 0);
double weight = item.edge(i)->get_data().up_weight+ item.edge(i)->get_data().down_weight;
if (weight == 0)
logstream(LOG_FATAL)<<"Bug: found zero edge weight between: " << item.id()-M+input_file_offset << " -> " << other_item-M+input_file_offset <<std::endl;
if (weight <= 1){
if (debug)
std::cout<<"skipping edge to " << item.id()-M+1 << " -> " << other_item-M+1 << " because of similarity is smaller or equal to one: " << weight << std::endl;
continue;
}
pivot_edges.mymutex.lock();
//add weight according to equation (15) in the probabalistic item similarity paper
set_val(pivot_edges.ratings, other_item-M, get_val(pivot_edges.ratings, other_item-M) + ((user_item_edge_weight-0.5)/0.5)* (weight- 1));
if (debug){
std::cout<<"Adding weight: " << (((user_item_edge_weight-0.5)/0.5)* (weight- 1)) << " to item: " << other_item-M+1 << " for user: "******" weight-1: " << weight-1<<std::endl;
std::cout<<pivot_edges.ratings<<std::endl;
}
pivot_edges.mymutex.unlock();
}
if (debug)
std::cout<<"Finished user pivot " << user_pivot << std::endl;
return 0;
}
/**
* add weighted ratings for each linked item
*
*/
double compute_ratings(graphchi_vertex<uint32_t, edge_data> &item, vid_t user_pivot, float edge_weight) {
assert(is_pivot(user_pivot));
if (!allow_zeros)
assert(edge_weight != 0);
else if (edge_weight == 0) {
zero_edges++;
return 0;
}
dense_adj &pivot_edges = adjs[user_pivot - pivot_st];
if (!get_val(pivot_edges.edges, item.id())) {
if (debug)
Rcpp::Rcerr<<"Skipping item pivot pair since not connected!" << item.id() << std::endl;
return 0;
}
int num_edges = item.num_edges();
if (debug)
Rcpp::Rcerr<<"Found " << num_edges << " edges from item : " << item.id() << std::endl;
//if there are not enough neighboring user nodes to those two items there is no need
//to actually count the intersection
if (num_edges < min_allowed_intersection || nnz(pivot_edges.edges) < min_allowed_intersection) {
if (debug)
Rcpp::Rcerr<<"skipping item pivot pair since < min_allowed_intersection" << std::endl;
return 0;
}
for(int i=0; i < num_edges; i++) {
vid_t other_item = item.edge(i)->vertex_id();
assert(other_item - M >= 0);
bool up = item.id() < other_item;
if (debug)
Rcpp::Rcerr<<"Checking now edge: " << other_item << std::endl;
if (is_user(other_item)) {
if (debug)
Rcpp::Rcerr<<"skipping edge to user " << other_item << std::endl;
continue;
}
if (!undirected && ((up && item.edge(i)->get_data().up_weight == 0) ||
(!up && item.edge(i)->get_data().down_weight == 0))) {
if (debug)
Rcpp::Rcerr<<"skipping edge with wrong direction to " << other_item << std::endl;
continue;
}
if (get_val(pivot_edges.edges, other_item)) {
if (debug)
Rcpp::Rcerr<<"skipping edge to " << other_item << " because alrteady connected to pivot" << std::endl;
continue;
}
assert(get_val(pivot_edges.edges, item.id()) != 0);
float weight = std::max(item.edge(i)->get_data().down_weight, item.edge(i)->get_data().up_weight);
if (!allow_zeros)
assert(weight != 0);
else if (weight == 0) continue;
pivot_edges.mymutex.lock();
set_val(pivot_edges.ratings, other_item-M, get_val(pivot_edges.ratings, other_item-M) + edge_weight * pow(weight,Q));
pivot_edges.mymutex.unlock();
if (debug)
Rcpp::Rcerr<<"Adding weight: " << weight << " to item: " << other_item-M+1 << " for user: "******"Finished user pivot " << user_pivot << std::endl;
return 0;
}