Skip to content
Prof Rod Avatar
The research map

Ruslan Salakhutdinov

UPMC Professor of Computer Science · Machine Learning Department — Core Faculty · Carnegie Mellon University

Back to the research map

At Prof Rod’s reading desk

Questions to explore

  • What makes inference in a model with several layers of hidden variables expensive?
  • What does a recognition model approximate, and where can it fail?

Research interests

Ruslan Salakhutdinov studies deep learning and representations, including generative models.

  • Deep learning
  • Generative models
  • Representation learning

Explore the work

Prof Rod’s reading notes

Start with a paper

  • AISTATS · 2010

    Efficient Learning of Deep Boltzmann Machines

    Prof Rod’s reading note: Compare the generative model with the recognition network that initializes inference. Ask which computation becomes cheaper and which approximation remains.

Bring a question of your own

Turn reading into research

Draw the generative and recognition paths. Separate the quantity the model defines from the approximation used to compute it.

Courses and teaching

Keep exploring

More at Carnegie Mellon University

Explore the university

Improve this page

Suggest a correction, share a paper to start with, or point readers to a group page or opening.