Fig. 3

The structure of the proposed model. The proposed model mainly uses embedding layer and convolutional layers. All sequences are encoded into numerical vectors of length 200 and are fed into the embedding layer. Each embedding vector dimension is 128. Then the outputs of embedding layer are fed into N convolutional layers. Each convolutional layer uses 64 filter kernels. These outputs are connected to feed into a max pooling layer and outputs of the pooling layers are concatenated to fed into another max pooling layer. Finally the output will be fed into a fully connection layer and passed through a sigmoid function. The final output is in range [0,1] as the prediction of the input sequence