Extensions/generalisations

  • relax linearity restrictions
  • try other hidden states instead of predictor coefficients