CIN Con1d problems

[MarkovSc]

In paper， section 3.2 CIN analysis,

the number of weight at k-th layer is h_k * h_{k-1} *m.
Howerver in you code, you use the con1d which has the weight of h_k * h_{k-1} *m * d.

should we use the tf.nn.depthwise_conv1d but the con1d, hope your responses

[Leavingseason]

I couldn't find the function of 'depthwise_conv1d'. Considering that conv1d is actually implemented via reshaping the input and then send it to conv2d, may be what you mentioned is exactly the current implementation of conv1d?

[MarkovSc]

In your version of CIN， the number of weight in k-th layer is field_nums[-1] * field_nums[0] *layer_size * dim。
The Con1d is the two-dimensional weight. But in the CIN analysis(section 3.2) , the CIN is independant of embedding dim which is just ( h_k * h_{k-1) * m in k-th layer)