Notes on tensorflow(一) Framework Overview
2017-04-27 17:41
501 查看
About tensorflow
Tensorflow自2015年11月被Google开源后迅速火了起来。听一位在MS实习的师兄说, MS也用Google的tensorflow。它能火起来的原因应该有以下几个:1. Google公司的大力推广
2. 非常活跃的开源社区
3. 功能强大, 好用“易”用
4. 。。。
Overview
Tensorflow, 简称tf, 不仅仅是一个deep learning的框架, 它更是一个general的machine learning计算框架。 tf采用了符号式编程(Symbolic style),它的计算是通过计算图(Computation Graph)实现的, 支持自动求导。同一类型的框架还有theano, 一个始祖型的老牌框架。与之相对的另一种编程模式是是命令式编程(Imperative Style),如caffe。 MXNet则同时采用了两者。使用Symbolic Style的最大好处莫过于自动求导。与caffe相比,不用自己实现backward方向。与theano相比, tf的graph使用起来感觉更好, 因为tf用户不用像theano用户一样等漫长的function build过程。tf同时提供low-level和high-level的接口。前者更灵活,可实现细粒度的各种操作, 足以满足general的Machine Learning编程需求。后者使用起来更方便, (重复)代码量相比于前者大大降低。后者是对前者的封装。
tf里有一个特殊的包:
tf.contrib
A high-level API like tf.contrib.learn helps you manage data sets, estimators, training and inference. Note that a few of the high-level TensorFlow APIs–those whose method names contain contrib– are still in development. It is possible that some contrib methods will change or become obsolete in subsequent TensorFlow releases.
https://www.tensorflow.org/get_started/get_started
大概意思是它是一个高级包, 提供很多功能, 但在以后的版本里可能被废除。
相关文章推荐
- Notes on tensorflow(八)read tfrecords with slim
- Notes on tensorflow(六)variable_scope
- Notes on Tensorflow
- Notes on Tensorflow(四)Variables
- Notes on learning tensorflow
- Notes on tensorflow(二)Get started
- Notes on tensorflow(七)将数据集转换为TFRecord
- Notes on tensorflow(五)Tensor Ranks, Shapes, and Types
- hands on machine learning with sklearn and tensorflow 附录B-扩充整理 2关于获取数据
- 浅尝TensorFlow on Kubernetes
- Installing TensorFlow on Ubuntu or Windows
- Softmax on Digits Data with TensorFlow
- TensorFlowOnSpark
- chapter2 of OReilly.Hands-On.Machine.Learning.with.Scikit-Learn.and.TensorFlow
- 如何安装Spark & TensorflowOnSpark
- TensorflowOnSpark 介绍与搭建
- Note on tensorflow(三)Get startd on Tensorboard
- Installing TensorFlow on Mac OS X
- <TensorFlow> Notes
- [note] deep learning tensorflow lecture 2 notes 深度学习笔记 (2) 解决过拟合