Implement the sequence representation and DNN model in Keras in the 2018 ICLR paper - An image representation based convolutional network for DNA classification https://github.com/Doulrs/Hilbert-CNN
python hilbertCNN.py ./data/data.txt
The image of the HilbertCNN architecture is captured from the paper. All data files in the data folder are from the paper.
The performance hasn't been reproduced as it is described in the paper. There could be several reasons:
- Training hyperparameters
- Not totally same CNN architecture as described in the paper
Improvements:
- How does it make sense biologically?