Multi-thread-Factorization-Machine

1.A very efficient factorization machine c++ multi-thread implemention，using google downpour method
for linear part, FTRL optimization method is used,for interaction part ,adagrad with l2 reguralization is used , only support two-order interaction 2.in the train file , each line contains a sample, in format click \t impre \t feat_index1 \t val1 feat_index2 \t val2 ...
3.gcc4.8 surpport
4.support rank loss and logistic loss
5.support read local .gz file or from hdfs
6. config file feat_num & fm_index &　hdfs_conf must be provided
7. in 8-core machine,it can process 100 million samples (hundreds non zero features) one epoch in one hour