TensorFlow Wide & Deep: Advanced Classification the easy way
summary
Demonstrates a wide and deep model that mixes a linear “wide” memorization path with a DNN embedding-based “deep” path, built with TensorFlow canned estimors. Trains on the 1994 census (≈32k train / 16k eval), reports about 83% accuracy, and shows TensorBoard for weight distributions. Exports the trained graph as a SavedModel export (saved_model.pb + variables) and serves predictions via Google Cloud ML Engine REST endpoints; the combined model improves rare-case memorization while keeping generalization.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 2:25What is TensorFlow
- 4:57How to use TensorFlow
- 7:05Can Estimators
- 11:20Building an app
- 14:27Deep Neural Networks
- 17:36Wide Deep
- 19:43Data
- 21:35Typical structure
- 28:20Code
- 36:37Tensorboard
- 39:35Scale
- 42:19Model
- 45:10Prediction
- 45:59Local vs Cloud
- 47:30Questions
related talks
40mTensorFlow Wide & Deep: Advanced Classification the easy way
43mMachine Learning at Scale
01h 06mDeep Learning with TensorFlow
01h 07mDeep Learning with TensorFlow
57mTensorflow, deep learning and modern convolutional neural nets, without a PhD
35mDeconstructing Deep Learning
59mTensorFlow in Three Sentences
37mDeep Recommender Systems