Getting started with TensorFlow
There are multiple open source deep learning frameworks besides TensorFlow (https://www.tensorflow.org) that I could have chosen for this sample application.
Some of the most popular frameworks are as follows:
- PyTorch (http://pytorch.org)
- Caffee2 (https://caffe2.ai)
- MXNet (https://mxnet.apache.org)
- Keras (https://keras.io): A high-level neural network abstraction API capable of running other deep learning frameworks such as TensorFlow, CNTK (https://github.com/Microsoft/cntk), and Theano (https://github.com/Theano/Theano)
TensorFlow APIs are available in multiple languages: Python, C++, Java, Go, and, more recently, JavaScript. We can distinguish two categories of APIs: high level and low level, represented by this diagram:
TensorFlow high-level API architecture
To get started with the TensorFlow API, let's build a simple neural network that will learn the XOR transformation.
As a reminder, the XOR operator has only four training examples:
|
X |
Y |
Result... |
