Machine Learning with TensorFlow

How Google does ML

Google infuses Machine Learning into almost all its product. There are also many pre-trained machine learning services available on Google Cloud.

Data is the fuel for ML models. ML converts examples into knowledge.

There are two stages of ML:

1. Train a ML model (a mathematical function) with examples (label + input): make tiny adjustments to model function so output is closer to label for a given input.
2. Run inference with a trained model

Data scientists must focus on both the training and inference stages of ML.

Training deep neural networks required computing power. For machine learning, you need to build a streaming pipeline in addition to a batch pipeline.

AI Platform Notebooks

Increasingly, data analysis and ML are carried out in self-descriptive, shareable, executable notebooks.

TensorFlow Playground

TensorFlow Playground is a powerful tool for visualizing how neural networks work.

ML with TensorFlow

What is TensorFlow

TensorFlow is an open-source high-performance library for numerical computation that uses direct graphs.

A DAG consists of tensors and operations on those tensors. A tensor is an N-dimensional array of data.

TensorFlow is popular among both deep learning researchers and machine learning engineers.

TensorFLow API

The Core TensorFlow Python API gives users full control to build and run.

tf.layers, tf.losses, tf.metrics provide components useful when building custom NN models.

tf.estimator is the high-level API for distributed training.

Lazy Evaluation

TensorFlow does lazy evaluation. You write a DAG, the directed acyclic graph, and then you run the DAG in the context of a session to get results. The graph definition is separate from the training loop because this is a lazy evaluation model.

First step, create the graph:

c = tf.add(a, b)

Second step, run the graph:

session = tf.Session()
numpy_c = session.run(c, feed_dict= ...)

Graph and Session

Graphs can be processed, compiles, remotely executed, and assigned to devices.

import tensorflow as tf

x = tf.constant([3, 5, 7])
y = tf.constant([1, 2, 3])

with tf.Session() as sess:
  print sess.run(z)

Tensor and Variables

Tensors

A tensor is an N-dimensional array of data. When you create a tensor, you typically specify its shape.

Tensors can be sliced:

import tensorflow as tf

x = tf.constant([[3, 5, 7],[4, 6, 8]])
y = x[:, 1]

with tf.Session() as sess:
  print y.eval()

Tensors can be reshaped:

import tensorflow as tf

x = tf.constant([[3, 5, 7],[4, 6, 8]])
y = tf.reshape(x, [3, 2])

with tf.Session() as sess:
  print y.eval()

Variables

A variable is a tensor whose value is initialized and then typically changed as the program runs.

Debugging TensorFlow programs

Steps

1. Read error messages to understand the problem
2. Isolate the method in question
3. Call the problematic method with fake data

Common Problems

• Shape problems
• Data type problems