TensorFlow is a deep learning framework open sourced by Google at the end of 2015. Open source library address: Click to open the link
The framework is mainly based on the data flow graph for numerical calculation, and the nodes represent the connection between the data.
At present, TensorFlow can be used to build deep learning neural networks such as CNN, RNN, and LSTM. Next, I will share Xiaobai's learning path, hoohoo!
First of all, you must start by installing Tensorflow. At present, TensorFlow supports Python and C++ languages, and the big guys can write the SWIG interface by themselves.
The tutorial to install Python can be found on Baidu. It is recommended to download Python 3.6. Only with the compilation environment, you can choose Python's own shell, or you can choose an integrated environment with a large number of scientific packages such as pycharm and Anaconda. ,
The installation of the Windows version is introduced here (the Mac version is unsuccessful, Orz...) If you choose the built-in shell compilation environment, you first need to download the TensorFlow download address: Click to open the link (if you do not have pip installed, please download pip yourself)
Then use the Windows command line to install pip: use the python setup.py install command in the Python directory to install pip. After installation, type python -m pip install tensorflow command, wait for the installation to complete, you can import TensorFlow in the compiler!
This article mainly introduces the learning process of the TensorFlow framework. The first step is to understand some basic knowledge of the TensorFlow framework:
The first is that under the TensorFlow framework, the computing process of learning the network needs to be represented as a flow graph. Data flow is a common programming model for parallel computing, so why use a data flow graph and how to build a data flow graph for computation? Graph operations are performed by calling Session after building the data flow graph . The introduction in the TensorFlow Chinese translation manual in the development manual of Tencent Cloud-Cloud + Community is more detailed: click to open the link and read this article to have a deep understanding of the TensorFlow framework.
In the TensorFlow framework, tensor is used to represent data, and the detailed explanation of this part of maintaining state through variables can be referred to: Click to open the link
After you understand these basic concepts, you can actually fight the TensorFlow framework. In the next article, we will start from the "Hello World" level example in the TensorFlow world: building CNN to realize MNIST handwritten digit recognition. Interested friends can comment and communicate together. Roar!