
Home Institution: University of Michigan
Dustyn Hofer
PhD StudentMy research:
I work on Di-Higgs and Higgs-to-invisible searches with the ATLAS detector.
My expertise is:
Fake-tau background estimation in the HH->bbtautau channel and forward jet modelling in the VBF+MET H->invisible search.
A problem I’m grappling with:
Connecting a series of analysis packages (nTuple generation, parallel processing, ML toolkits) together into one consistent, reproducible analysis pipeline.
I’ve got my eyes on:
Ways to use our physics knowledge to ask the right questions for the right kinds of machine learning architectures to take best advantage of their properties.
I want to know more about:
How different ML architectures are more or less helpful for certain classes of problems.

