Lecturer in biomedical engineering at UC Berkeley, focusing on sustainable technology solutions.
Computational statisticians finalized regularized loss functions that allowed deep learning models to successfully train and learn patterns even when faced with highly corrupted, mislabeled training data sets. Part of the 34 AI Roots Facts: 2015… Read More »The Formulation of Robust Optimization for Deep Networks under Label Noise
The period between 2001 and 2005 taught computer science that data storage and processing power could no longer be contained within single server chassis. To handle the explosive growth of consumer internet activity, engineers had… Read More »The Transition to Cloud-Native Distributed Infrastructure
Systems engineers published mathematical optimization protocols for splitting coordinate descent tasks across massive compute clusters, enabling enterprise advertising networks to parse petabyte-scale ad-click matrices instantly. Part of the 32 AI Roots Facts: 2016 Edition archive.… Read More »The Formulation of Distributed Parallel Coordinates Descent for Massive Linear Models
The defining lesson of 2009 was that the decades-old search for the perfect, complex machine learning algorithm was a secondary pursuit. By creating ImageNet and proving that raw, hyper-scale data volume combined with brute-force GPU… Read More »The Ultimate Realization of the Data-First Paradigm