This chapter briefly introduces the installation and use of TensorFlow. Some details need to be grasped slowly in subsequent applications.
TensorFlow is not just limited to neural networks and machine learning, it can even be used for quantum physics simulations.
Advantages of TensorFlow:
- Can run on many operating systems
- Provides a simple Python API called TF.Learn (tensorflow.contrib.learn), compatible with Scikit-Learn, which can train a variety of neural networks in just a few lines of code.
- Provides another API called TF-slim (tensorflow.contrib.slim) to simplify the creation, training, and evaluation of neural networks.
- Other high-level APIs are independently created on top of TensorFlow, such as Keras and Pretty Tensor .
- The core Python API provided is flexible (at the cost of high complexity) and enables a wide variety of computations, including arbitrary neural network architectures.
- Contains C++ implementations that require efficient ML operations, and also provides C++ APIs to customize operations.
- Provides some advanced optimization algorithms to find parameters that minimize the loss function. The automatic differentiation algorithm is implemented , which can easily calculate the derivative of any function.
- Provides a visualization tool TensorBoard, which can be used to view calculation graphs, learning curves, etc.
- Google provides cloud service to run TensorFlow graphs .
- Resource pages: www.tensorflow.org , https://github.com/jtoy/awesome-tensorflow . Communication area: https://groups.google.com/a/tensorflow.org .
The remaining basic operations of TensorFlow will be familiar to you in the subsequent chapters.