Multiresolution Tensor Learning for Efficient and Interpretable Spatial Analysis
Offered By: Institute for Pure & Applied Mathematics (IPAM) via YouTube
Course Description
Overview
Explore efficient and interpretable spatial analysis techniques in this conference talk from the Tensor Methods and Emerging Applications workshop. Dive into the Multiresolution Tensor Learning (MRTL) algorithm, designed to overcome computational challenges in tensor latent factor models. Learn how MRTL improves interpretability and reduces computation by initializing latent factors from an approximate full-rank tensor model and progressively learning from coarse to fine resolutions. Discover the algorithm's theoretical convergence, computational complexity, and practical applications in basketball play modeling and precipitation forecasting. Gain insights into MRTL's 4-5x speedup compared to fixed resolution approaches while maintaining accuracy and interpretability in real-world datasets.
Syllabus
Intro
Modeling Basketball Play
Technical Challenges
Tensor Methods
Example: Profile Player Shots
Tensor Latent Factor Model
Example: Precipitation Forecast
Difficulty of Training
Multi-Resolution Learning
Rate of Convergence
Computational Complexity
Efficiency: MRTL is Fast
Sensitivity to fine-graining criteria
Interpretability: basketbacll
Interpretability: climate
Conclusion
Taught by
Institute for Pure & Applied Mathematics (IPAM)
Related Courses
Neural Networks for Machine LearningUniversity of Toronto via Coursera 機器學習技法 (Machine Learning Techniques)
National Taiwan University via Coursera Machine Learning Capstone: An Intelligent Application with Deep Learning
University of Washington via Coursera Прикладные задачи анализа данных
Moscow Institute of Physics and Technology via Coursera Leading Ambitious Teaching and Learning
Microsoft via edX