YIP: Learning, Dynamics and Intervention in Large-Scale Social Networks
Abstract
What is the mathematical objective of your project? What question are you trying to answer? Today we are facing a data deluge in almost every domain. Online social networks have seen an explosion in activity and have fundamentally transformed the nature of human interaction. In the biological realm, modern genome sequencers can output data at a rate 400 times faster than the ones a decade ago, and so on. However, although having a transformative potential, the data deluge has not yet been exploited to the fullest extent. Ironically, the data deluge has also resulted in a data desert. The collected data in many domains are noisy, subsampled, with typically a large number of variables or unknowns compared to the number of observations or the knowns. Such high-dimensionality entails practical principled approaches for learning from ill-posed and ill-behaved data. Some of the fundamental questions in high-dimensional learning are: Can we design scalable models for efficiently representing and learning high-dimensional data? Here, scalability refers to low computational requirements and reduced sampling of high-dimensional data. Not all phenomena can be learnt in a scalable manner. Can we characterize the fundamental limits on complexity of learning complex phenomena?
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 30, 2016
- Accession Number
- AD1064387
Entities
People
- Anima Anandkumar
Organizations
- University of California, Irvine