LIN 389C: Topic list Fall 2018

The consensus was that we should have several main themes. The following themes were mentioned:

  • Neural models: Autoencoders, variational autoencoders.
    And let me add: graph embeddings.
    Generative models were also mentioned -- was that in this context?
  • "Thinking, fast and slow": Probabilistic programming languages
  • Working with less data: semi-supervised learning, domain adaptation, distant supervision
  • Applications:
    • Multimodality
    • Entity-based modeling
Additional applications not mentioned in our discussion but that I think we should add: summarization, narrative modeling.

We will choose from this topic list in our first meeting after the break.

We have 13 weeks this semester in which we can do readings. Suggested reading order:
  • 2 weeks on an application: multimodality, narrative modeling, summarization -- your choice
  • 2 weeks on autoencoders (advanced applications; we did the basics of this already)
  • 3-4 weeks on graph embeddings
  • 2 weeks on probabilistic programming languages (mostly lecture, by Katrin)
  • 3 weeks on working with less data