Model structure of CNN_L3_wide.

CNN_L3_wide consists of 3 layers but with a wider configuration, meaning each layer has more filters. The wider structure allows for a broader capture of features at each level, facilitating the recognition of diverse patterns within the input data. This model also operates with a learning rate of 0.0001. #features stands for the number of input features. For instance, if the embedding method used is One-Hot, the #feature=21. These different configurations highlight the exploration of depth versus breadth in CNN architectures, providing insights into the trade-offs between layer depth and layer width in capturing features from protein sequences.
(XLSX)