LorenzoRimella/HiddenMarkovNeuralNetwork

An hybrid between HMM and Neural networks for sequential training and time series forecasting

HiddenMarkovNeuralNetwork

An hybrid between HMM and Neural networks for sequential training and time series forecasting. In this repository we display some tutorials on how to use HMNN in different scenarios.

Requirements for the tutorials

• FLAG-example. In "FLAG-example.ipynb" we explains how to run an HMNN for video frame prediction. You require the libraries:

  • cv2
  • torch (PyTorch)
  • random
  • sklearn
  • matplotlib
  • numpy
  • pickle: this is use only for saving and loading files

• MNIST_evolvingClassifier-example. In "MNIST_evolving_classifier-example.ipynb" we explains how to use HMNN with an evolving classifier on MNIST. We also show how to build the dataset. You require the following libraries:

  • gzip
  • torch (PyTorch)
  • random
  • sklearn
  • matplotlib
  • numpy
  • pickle: this is use only for saving and loading files

• MNIST-variationalDropConnect. In "MNIST_variational_dropconnect-example.ipynb" we explains how to use HMNN with Variational DropConnect in MNIST. You require the following libraries:

  • gzip
  • torch (PyTorch)
  • random
  • sklearn
  • matplotlib
  • numpy
  • pickle: this is use only for saving and loading files