inline void setup_node(db::node *node)
{
vm::predicate *init_pred(vm::theProgram->get_init_predicate());
vm::tuple *init_tuple(vm::tuple::create(init_pred, &(node->alloc)));
node->set_owner(this);
node->add_linear_fact(init_tuple, init_pred);
node->unprocessed_facts = true;
}
void random_forest_p(const data_t& train, const vec_data_t& test, preds_t& train_preds, vec_preds_t& test_preds, args_t& args) {
int trees = args.trees;
int numthreads=args.processors, i, x, t;
int fk = args.kfeatures;
if (numthreads > trees)
numthreads = trees;
trees = (trees / numthreads) * numthreads;
int trees_per_thread = trees / numthreads;
thread** threads = new thread*[numthreads];
vector< vec_preds_t > seg;
vector< preds_t > train_seg;
for (i=0; i < numthreads; i++) {
vec_preds_t p;
init_pred_vec(test, p);
seg.push_back(p);
preds_t tp;
init_pred(train, tp);
train_seg.push_back(tp);
}
for (i=0;i<numthreads;i++)
threads[i] = new thread(bind(multiple_forest, trees_per_thread, cref(train), cref(test), ref(train_seg[i]), ref(seg[i]), ref(args)));
for (i=0;i<numthreads;i++){
threads[i]->join();
delete threads[i];
}
fprintf(stderr, "done threading\n");
delete[] threads;
// congregate results of the threads
for (i = 0; i < numthreads; i++)
for (t = 0; t < test.size(); t++)
for (int j = 0; j < test[t].size(); j++)
test_preds[t][j] += seg[i][t][j];
for (i = 0; i < numthreads; i++)
for (int j = 0; j < train.size(); j++)
train_preds[j] += train_seg[i][j];
// average results of the trees
for (i = 0; i < test.size(); i++)
for (int j = 0; j < test[i].size(); j++)
test_preds[i][j] /= trees;
for (int j = 0; j < train.size(); j++)
train_preds[j] /= trees;
// write results to file
for (i = 0; i < test.size(); i++) {
tuple::write_to_file(test_preds[i], test[i], args.test_outs[i]);
}
}
