Machine Learning at Scale

on YouTube ↗43m

" summary

Builds a TensorFlow canned estimator (a wide-and-deep model) and trains it on the 1994 US census sample of ~32,000 rows locally (batch size 40, 1,000 steps, DNN hidden layers 100 and 50). Exports a SavedModel.pb with variables, uploads it to Google Cloud Storage and deploys a version to Cloud ML Engine; online REST predictions and batch jobs work and return probabilities. Notes auto-scaling behavior, warm‑up latency, and rough pricing (~$0.10 per 1,000 requests plus ~$0.40 per node hour).

AI-written summary · about

" chapters (5)

1. 0:00Start
2. 7:02create a linear model
3. 19:24mapping the type between strings and floats
4. 31:24using the prediction service for your model for images
5. 36:07loading a model into a mobile device

" related talks

• 37mEnd-2-End Machine Learning on GCPVikram Tiwari · DevFest Florida 2019
• 45mServerless Machine Learning: a beginner's guideBret McGowen, Yufeng Guo · Devoxx Belgium 2017
• 46mMachine Learning with Scikit-Learn and Xgboost on Google Cloud PlatformSteve Greenberg · Google Cloud Next '18
• 37mWhich Serverless Prediction Platforms on Google Cloud Platform?Guillaume Blaquiere · DataXDay 2020
• 32mMachine Learning with Google Cloud PlatformTiago Henriques · DSC Europe 2022
• 51mTensorFlow Wide & Deep: Advanced Classification the easy wayYufeng Guo · DevFest Nantes 2017
• 59mMachine Intelligence at Google ScaleGuillaume Laforge · Voxxed Days Singapore 2017
• 13mTaking Machine Learning to the Next LevelGiovanni Galloro · Codemotion Rome 2018