Modern Paradigms in Generalization Boot Camp

The bootcamp will bring participants up to speed on existing topics and open problems in generalization.  Due to the speed of the field, details will be clarified in the months leading to the workshop, but will include not just classical but also modern generalization settings such as robustness and out-of-distribution prediction; consequences of large models for vision and language tasks; optimization and algorithmic consequences; consequences of fine-tuning, RLHF and many other algorithmic inventions outside the standard one-pass ERM setup.  Lastly, the workshop will present some initial open problems ranging across not just standard ML tasks such as vision and language, but also the sciences.

If you require special accommodation, please contact our access coordinator at simonsevents [at] berkeley.edu with as much advance notice as possible.