Where are you headed for Summer vacation? You might pack your flip flops and head to the beach, or you might dig your boots out of the closet and explore a new hiking route. Qwiklabs is dreaming of out-of-office experiences too, with a brand new Scientific Data Processing Quest full of labs designed to transport you from your desk, to the coast, to NYC, even to the tropics. There’s room for Google Cloud training in your carry-on!
The Scientific Data Processing Quest walks you through using the Google Cloud Platform (GCP) to tackle complex data analyses, with minimal IT know-how (and minimal cost!). Think training a machine learning model is outside your skillset? It’s not! There are 7 labs in this Quest and each one walks you through using basic GCP tools to process huge datasets, and return fascinating results.
Here’s some of what you’ll do:
Build an earthquake prediction model to help you choose where NOT to vacation next Summer. (Or maybe you like experiencing earthquakes – that’s fine too!)
Use BigQuery to process and correlate two extremely diverse datasets, weather and civil complaints in NYC. Can you think of other datasets you might want to correlate?
Use Google Cloud Dataflow to build a monthly vegetation growth index based on Landsat images. You’ll learn to use Google Cloud Dataflow, a framework that automatically distributes work to many machines. You will work with an agriculture use case, but you can apply your new skills across industries. Virtually trek through Reunion Island while you’re at it. (link to Lak’s post from Nov)
Use transfer learning to train a machine learning model to classify coastline images. What’s transfer learning? A shortcut for artificial intelligence:
An image classifier consists of two parts: a feature extraction part that is many layers deep and a classifier part that is quite simple. In transfer learning, when you build a new model to classify your dataset, you reuse the feature extraction part from an already trained image classifier and re-train the classification part with your dataset. Since you don’t have to train the feature extraction part (which is the most complex part of the model), you can train the model far more quickly with fewer images and less computational resources.
– Lak Lakshmanan, Tech Lead, Data and ML Professional Services, Google Cloud
Don’t forget sunscreen!