function _build_extreme_FM_quick in exDeepFM.py file

[liuzhiyong01]

hparams.logger.info("split connect")
if idx != len(hparams.cross_layer_sizes) - 1:
next_hidden, direct_connect = tf.split(curr_out, 2 * [int(layer_size / 2)], 1)
final_len += int(layer_size / 2)
else:
direct_connect = curr_out
next_hidden = 0
final_len += layer_size
field_nums.append(int(layer_size / 2))

I want to know why does "next_hidden, direct_connect = tf.split(curr_out, 2 * [int(layer_size / 2)], 1)" this action

[Leavingseason]

It's a little trick. We found empirically that the performance is better via splitting direct connections (to output unit) and hidden layers to next stage.