Engineering Approximate Computations

There’s a new ecosystem of applications that integrates machine learning into a variety of tasks. Typical domains have included image recognition and natural language processing. However, these techniques have also spread to computer systems domains, such as program compilation, resource scheduling, and database query optimization, yielding new computer systems that learn from data to achieve their goals.

With the success of these systems, we must grapple with the reality that they model and compute with objects that are inherently approximate — real numbers (only computable up to a given precision), neural networks (only validated on a given dataset), and probabilistic computations (results only computable up to a given probability). This reality presents many engineering questions about interpreting, debugging, validating, verifying, and optimizing these systems.

As an illustrative example of such a system, I'll present our work on Ithemal and DiffTune, our deep learning systems for learning the behaviors of modern computer processors. Together these systems combine parameterized programs with neural networks together to learn interpretable, neurosymbolic models.

Guided by these engineering challenges, I’ll also discuss our work on the semantics of and reasoning for sound real-valued, differentiable, probabilistic computation, which is the core computational model behind this new class of systems.

----

Please register here to join us for the CS Colloquium Series via Zoom this fall. Once you register, you will receive the Zoom link. You only need to register once to be able to attend any of the seminars.