Extending Dynamic Time Warping to Directed Graphs and DAGs for Improved Pattern Recognition and Alignment
Organization
ITS Lab
Abstract
Extending Dynamic Time Warping to Directed Graphs and DAGs for Improved Pattern Recognition and Alignment
This thesis proposal addresses the need for advanced pattern recognition and alignment techniques in domains such as image recognition, map matching, and multi-sequence alignment. Current methods struggle to capture complex relationships within directed graphs and Directed Acyclic Graphs (DAGs). To bridge this gap, the proposal aims to extend the Dynamic Time Warping (DTW) technique for handling these structures, resulting in more accurate pattern recognition.
- Research Objectives:
1. Develop algorithms to adapt DTW for directed graphs.
2. Extend the proposed model to work with Directed Acyclic Graphs.
3. Design a suitable training process for the extended model.
4. Compare the performance of the proposed model with Hidden Markov Models (HMMs) in various application domains.
5. Compare with image template matching techniques.
- Training and Learning:
1. Utilize supervised learning techniques with relevant datasets for each application domain.
2. Focus on optimizing model-specific parameters for enhanced performance.
- Comparison with HMMs:
Benchmark the proposed model against HMMs in terms of accuracy, robustness, and efficiency.
This thesis proposal addresses the need for advanced pattern recognition and alignment techniques in domains such as image recognition, map matching, and multi-sequence alignment. Current methods struggle to capture complex relationships within directed graphs and Directed Acyclic Graphs (DAGs). To bridge this gap, the proposal aims to extend the Dynamic Time Warping (DTW) technique for handling these structures, resulting in more accurate pattern recognition.
- Research Objectives:
1. Develop algorithms to adapt DTW for directed graphs.
2. Extend the proposed model to work with Directed Acyclic Graphs.
3. Design a suitable training process for the extended model.
4. Compare the performance of the proposed model with Hidden Markov Models (HMMs) in various application domains.
5. Compare with image template matching techniques.
- Training and Learning:
1. Utilize supervised learning techniques with relevant datasets for each application domain.
2. Focus on optimizing model-specific parameters for enhanced performance.
- Comparison with HMMs:
Benchmark the proposed model against HMMs in terms of accuracy, robustness, and efficiency.
Graduation Theses defence year
2023-2024
Supervisor
Kaveh Khoshkhah
Spoken language (s)
English
Requirements for candidates
Level
Masters
Keywords
Application of contact
Name
Kaveh Khoshkhah
Phone
E-mail