Fig. 1.

General illustration of our solution. We represent an RNA sequence as two-dimensional map with 8 channels with one-hot encoding (each color in the input tensor represents a one-hot vector of size 8. The gray box shows what each color means. for example, dark green is a potential GC pairing which is [0,0,0,0,0,0,1,0]). We process the map with a convolutional network which produces a score matrix for all possible pairings. Finally, we convert the score matrix into the RNA secondary structure