Serverless Machine Learning: An Introduction
summary
A live pipeline demonstrates ingesting 600–700 handwritten sevens eights and nines into Cloud Storage, transforming each image into 784 grayscale pixel columns stored in BigQuery, triggering Cloud Functions → BigQuery → Cloud ML Engine to retrain a TensorFlow model. The initial model misclassifies sevens, training on the new batch raises measured accuracy from about 92% to about 96%, the updated model deploys and predicts correctly, and AutoML is recommended for image cases when available.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:24Traditional Image Detection
- 5:12How Machine Learning Works
- 9:16Training New Models
- 9:28Adding Files
- 9:46AutoML
- 10:43How ML Works
- 11:19Serverless ML Code
- 12:08TensorFlow
- 21:53Cloud Functions
- 27:58Cloud ML Engine
- 28:48Failed Demo
- 29:44Cloud AI Adventures
- 34:38Sanity Check
- 35:38Data Prep
related talks
45mServerless Machine Learning: a beginner's guide
57mBuilding smarter solutions with Machine Learning, from magic to reality
38mReal-world Machine Learning with TensorFlow and Cloud AI
32mMachine Learning for Game Developers
22mTrain Custom Machine Learning Models with No Data Science Expertise
40mServerless Machine Learning 101
41mBuilding Serverless Machine Learning Models in the Cloud
54mBuilding smarter apps with Machine Learning, from magic to reality