Jordan Brown

I study learning theory and mathematical logic at the University of California, Berkeley, in the Group for Logic and the Methodology of Science. Thomas Scanlon and Antonio Montalbán advise me.

I am a Ph.D. candidate. I mostly work on private learning theory. The intersection of statistical mechanics and learning interests me. And I am trying to understand how transformer weight assignments are selected, how such selection resembles the natural selection of bio-organisms, and what such selection implies for the psychology of neural networks.

Published and Unpublished Works

Private Learning on Linear Orders (I learned after writing this that its primary results were known and the questions I ask in it have been answered. But the use of the logistic function is, as far as I know, novel.)

Definability of the Integrability Locus. This was also known through work of Valette. The approach I used required untangling some of the implications of valuative Lipschitz stratifications for classical Lipschitz stratifications; T. Kaiser told me of Valette's work.

When and Why to Expect Gaussian Error Bars. Joint work with M. Wilensky and B. Hazelton on statistical problems in astronomy.