Caffe is a popular machine vision deep learning tool. We've made our own extensions to Caffe 1.0 which are useful for timeseries prediction:

  1. Gaussian Noise data augmentation layer.
  2. Early Stopping Solver.
  3. Gaussian Data noise layer.
  4. Roll utility layer.
  5. LSTM bias filler.
  6. LSTM Data Windowing Layer.
  7. General Purpose Data Windowing.
  8. "Load Once" HDF5 data layer.
  9. Moving Average data layer.
  10. Input Scaling data layer.
  11. Input Normalization data layer.
  12. Input Differencing data layer.

Feel free to use if you are interested! If you have any questions or suggestions, connect with us at We'll do our best to help.