The 18th European Conference on Machine Learning (ECML) and the 11th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD)

Separating Precision and Mean in Dirichlet-enhanced High-order Markov Models

author: Rikiya Takahashi, IBM Research

Categories

Top: Computer Science: Machine Learning: Markov Processes

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Separating Precision and Mean in Dirichlet-Enhanced High-Order Markov Models
0:09	Agenda - Necessity of robustly estimating high-order Markov process models
0:54	Necessity of robustly estimating high-order Markov process models - Natural language and Markov models
1:47	Necessity of robustly estimating high-order Markov process models - Problem caused by data sparseness
2:36	Necessity of robustly estimating high-order Markov process models - Introducing smoothing methods
3:12	Agenda - Prior work: estimating Markov process models by hierarchical Bayesian approaches
3:16	Prior work - Two major smoothing criteria
4:46	Prior work - Smoothing methods = Hierarchical Bayesian estimation
6:05	Prior work - Known performances of existing methods
7:33	Prior work - Frequency modification by an indicator function
8:53	Agenda - Our proposition: Separating precision and mean in Dirichlet prior
8:56	Our proposition - Our direction
10:01	Our proposition - Discounting factor should depend on current states.
12:09	Our proposition - Separating precision and mean in Dirichlet prior
12:44	Our proposition - New formulation : context-dependent Dirichet prior
13:08	Our proposition - Effective frequency for more precise lower-order distribution
14:16	Our proposition - New Dirichlet prior will outperform when # of states is small.
14:57	Agenda - Experimental result
14:59	Experimental result - Checking the performances depending on the # of states.
15:34	Experimental result : evaluating test-set perplexity - Natural language modeling : slightly worse than Kneser-Ney smoothing
16:08	Experimental result : evaluating test-set perplexity - Protein sequence modeling : outperformed Kneser-Ney smoothing (1)
16:51	Experimental result : evaluating test-set perplexity - Protein sequence modeling : outperformed Kneser-Ney smoothing (2)
16:59	Agenda - Conclusion
17:01	Conclusion

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.