Concurrency Is Not Parallelism
Clarifies concurrency as the composition of independently executing processes (Hoare's CSP) versus parallelism as simultaneous execution, illustrated with live worker-pool topologies.
Faculty & visiting speakers
Talks & colloquia from Google
7Self-contained seminars, public lectures, and invited talks.
Cluster Computing and MapReduce: Introduction (Lecture 1)
This inaugural lecture introduces fundamental concepts of cluster computing and the MapReduce programming model, detailing its architecture and fault tolerance mechanisms for large-scale data processing.
Leveraging Per-Instance Privacy for Machine Unlearning
Anvith Thudi presents a principled, per-instance privacy framework for quantifying machine unlearning difficulty, offering improved utility-unlearning tradeoffs.
POPRi: Private Federated Learning using Preference-Optimized Synthetic Data
Charlie Hou details POPRi, an innovative approach to private federated learning that leverages preference-optimized synthetic data to enhance model training while preserving privacy.
Hacking the Kernel with Go: A Practical Introduction to eBPF
An introductory yet comprehensive exploration of eBPF, demonstrating how to leverage Go for secure, in-kernel programming without modifying the Linux source.
Dual Averaging Method for Regularized Stochastic Learning and Online Optimization
A presentation on the dual averaging method for regularized stochastic learning and online optimization, recognized with the NeurIPS Test of Time Award for its foundational impact.
From Unicorns to Centaurs: Powering the AI Moonshot
Partha Ranganathan, a Google Technical Fellow, discusses the architectural shifts required to power the next generation of AI systems, drawing from his extensive work on datacenter design.