Description
Machine learning models of social media popularity abound, but ones that take into account finite attention or competition among similar memes are scarce. Feature-driven approaches try to learn a function that maps content, user, and network features directly to an output variable indicating popularity, such as the volume, popularity rank, or change in popularity. Another class of models explicitly describe each sharing event as they happen over time. The mathematical tools used arise from stochastic point processes, such as self-exciting processes, cox survival processes, or stochastic epidemic models. The key to our proposed solution lies in incorporating work on choices in economics with models of online attention. We propose three subtasks to analyse and design models for the competition dynamics of memes. We will create novel methods to characterize the attention market in social media. This will extend the stochastic process model with a choice model component, provide better market models that take into account item quality and extrinsic environment, and study the gap between individual choice and social choices. We will formulate the main model form of the choice process model and explore several estimation algorithms. One contribution will be designing a choice modulator function to be able to take into account network capacity constraints.
Besides making foundational contributions in machine learning algorithms, the broader implication includes addressing the challenge of manipulatable online attention that threatens our democracy, as well as potential applications in other domains including the evolution of organisms and ecosystems. The student is expected to develop the research and methods for the above problems, publish and present the corresponding outcomes at top machine learning venues (such as NeurIPS, ICML, ICLR, AISTATS) and application venues. They will also be given the opportunity to work alongside our collaborators at the University of Warwick (UK).
Requirements
- Degree in computer science, statistics or related quantitative fields
- Excellent knowledge of machine learning techniques (e.g. popular supervised and unsupervised learning methods and fundamental concepts such as generalization, regularization, overfitting)
- Expertise in high-level programming languages such as Python
- Desirable: Knowledge of probabilistic inference techniques and deep learning frameworks such as Pytorch and TensorFlow
Collaborations
This project will be carried out in collaboration with Prof. Lexing Xie from ANU.
