Large-Scale Machine Learning for Urban Planning
• October 5, 2020
Brett Naul, founding engineer at Replica, joins Matt Rocklin and Hugo Bowne-Anderson to discuss large-scale machine learning and travel simulations for urban planning. Replica uses Dask to easily scale travel simulations to hundreds of millions of agents on Google Container Engine.
The rich Python data science and statistical ecosystems make it easy to build new representations of human movement and activity. Replica uses Dask to scale models that make use of many different libraries, most of which have no built-in Dask integration but are still easy to parallelize using a simple set of Dask patterns. We also use the same cloud infrastructure to scale more standard data science analyses using numpy, pandas, and xarray with no additional overhead.
After attending, you’ll know
- How probabilistic graphical models can help build a privacy-preserving representation of a population,
- How Helm and Kubernetes can be used to deploy Dask alongside custom microservices, and
- How Dask and Google BigQuery can be used together to tackle petabyte-scale datasets.
Join us this Thursday, October 8th at 5pm US Eastern time by signing up here and dive into the wonderful world of large-scale machine learning for urban planning!