Markov Models as compression models of interesting sequences

 

• Simple Markov models
  • Zero-offset
  • Neighbor blurring
  • Complement blurring
• Hidden Markov Models
• Strengths and weaknesses of hidden Markov models
• Violations of assumption needed for cost maps
• Building a hidden Markov model from a simple Markov model
• Direct construction of HMM from a sequence
• Training a hidden Markov model
  • Tuning and pruning
  • Cross-training
• Modifying an HMM for better matches to a seed
  • RemoveUseless
  • MergeStates
  • AddUseful
  • AddSkipEdges
  • UnrollSelfLoops
  • SplitVagueStates