Navigating TensorFlow Estimator Documentation

Sunday May 21, 2017

The TensorFlow documentation keeps improving, but it can still be hard to find what you're looking for. Here's a way of organizing Estimators or tf.contrib.learn documentation, from high-level to low-level:

• Using Estimators (already made or "canned")
  • DNNClassifier on iris dataset
  • LinearClassifier (logistic) on census income dataset
  • DNNLinearCombinedClassifier (extends census income example)
• Aspects of using Estimators
  • Feature columns (in context of linear models)
  • Using input_fn; DNNRegressor on Boston housing dataset
  • Logging etc. with ValidationMonitor (extends iris example)
• Making your own Estimator
  • With layers in a model_fn on the abalones dataset
  • With layers in a model_fn on MNIST
• Lower-level API than Estimators
  • Use the lower-level API to make a system for MNIST