Related Experiment Videos
A learning method of hidden Markov models for sequence discrimination
1C&C Research Labs., NEC Corporation, Kanagawa Japan.
Summary
We developed a new learning method for hidden Markov models (HMM) that significantly reduces sequence discrimination errors. This approach effectively utilizes negative examples to improve HMM training for identifying unknown sequences.
Area of Science:
- Machine Learning
- Pattern Recognition
- Computational Statistics
Background:
- Hidden Markov Models (HMMs) are widely used for sequence analysis.
- Traditional HMM training often relies solely on positive examples, potentially limiting discrimination accuracy.
- Sequence discrimination is crucial in various fields, including bioinformatics and speech recognition.
Purpose of the Study:
- To propose a novel learning method for HMMs specifically designed for enhanced sequence discrimination.
- To incorporate the use of negative examples into the HMM training process.
- To demonstrate the effectiveness of the proposed method compared to existing techniques.
Main Methods:
- A gradient descent algorithm is employed to train HMM parameters.
- A cost function is defined as the product of likelihood differences between observed and desired sequences.
- The method integrates both positive and negative training sequences.
Main Results:
- The proposed HMM learning method significantly reduced discrimination errors compared to two existing methods.
- Experimental evaluations using cross-validation demonstrated superior performance.
- The inclusion of negative examples proved beneficial for HMM training.
Conclusions:
- The developed learning method offers a robust approach for sequence discrimination using HMMs.
- The strategic use of negative examples is a key factor in improving HMM performance for distinguishing unknown sequences.
- This method provides a valuable tool for applications requiring accurate sequence classification.