The operation on two modal feature

[yuxinyan]

HI, i found that in your paper you use element-wise add, multiplication and vector concatenation followed by a fc layer, but you only use element-wise add, multiplication and vector concatenation followed without a fc layer in your main code. So the output dimension is 4* ds not 3*ds described in sec3.2. Is there a mistake ?