### Research Interests

I’m working on a variety of topics, loosely collected under the heading of finding or creating useful structure in neural representations

• how can we better understand the layer-wise steps that neural networks take transforming inputs to ouputs?
• what are the advantages of a “modular” neural network, and how do we build them?
• how does the brain learn and model causal relationships?

And growing out of my earlier work on Bayesian inference in the brain, a separate line of my recent work asks

• what sort of approximate inference algorithms exist “between” MCMC and variational inference?
• (how) does the brain represent probability?

### Bio

My undergrad was at Dartmouth College, where I mostly did Computer Science and Engineering, but sparked an interest in the connection between AI and neuroscience. This led me in 2014 to a PhD program in Computer Science at the University of Rochester, where I quickly discovered that making “brain inspired AI” means first understanding “brains.” I transferred to the Brain and Cognitive Science department in 2015, where I did my main PhD work on Bayesian Inference in low-level visual perception, graduating in fall 2020.

By the end of my PhD, I saw some serious flaws in the Bayesian framework as a tool for understanding (and building) neural computation. This has led down three paths in my postdoc work:

• expanding what is meant by “Bayesian inference” by developing new algorithms
• diving into the philosophy of what “neural computation” is in the first place
• turning my attention to deep learning and asking what kind of useful structure is there, or could be built in

[lastname].[firstname].d@gmail.com