Skip to main content
. 2023 Feb 17;13:2861. doi: 10.1038/s41598-023-30124-x

Figure 3.

Figure 3

Model architecture. The input of our model is a one-hot matrix with size 5*L which encodes a subsequence of length L. This matrix is then transformed into a matrix with size C*L and U*L by the 1D-CNN and Bi-LSTM layers, respectively. In fact, every base in the input RNA is encode by U units. Then, every base, encode by U units, is transformed by the following FC layers, resulting in an output with a size 1*L.