Wang, Ruoxi, et al. "Deep & cross network for ad click predictions." Proceedings of the ADKDD'17. 2017. 1-7. input_shape = [batch_size, field

Factorization Machine models pairwise (order-2) feature interactions without linear term and bias. input shape = (batch_size,field_size,embedding_

A `Dataset` that balance input data between other concurrent ops

calculate average of label

calculate average of label

Calculate incremental count, sum. Calculate incremental (data - mean).sqrt() as var

One layer to return a one-hot tensor base on provided value collection. For each row, will return len(position_array) - 1 zeros, and a one lo

Save ps weight after every fit.

NOTE! do not include brpc/server.h in this header file, its will cause butil LOG macro conflict with tensorflow LOG macro