Click on " Program Life " above, and select "Public Account"
Pay attention to the stories around the programmer (Yuan) for the first time
What is TensorFlow?
The predecessor of TensorFlow was DistBelief developed by the Google Brain team. Since its creation, it has been used by dozens of teams on countless projects, including deep neural networks. However, like many pioneering engineering projects, DistBelief has some design errors that limit its ease of use and flexibility. Later, Google launched a new project, which is TensorFlow, which is the most popular open source deep learning framework. It has rich applications in graphics classification, audio processing, recommendation systems, and natural language processing.
Although powerful, its framework learning threshold is not high, as long as you master the installation and use of Python, and have some knowledge of machine learning and neural networks, you can get started.
What is Theano?
Theano is a Python library that can run fast numerical computations on CPUs or GPUs. This is a key foundational library in Python deep learning, and you can use it directly to create deep learning models or wrapper libraries, greatly simplifying the program.
At the heart of Python, Theano is a compiler for mathematical expressions. It knows how to take your structure and make it a very efficient code using numpy, efficient native libraries like BLAS and native code (C++), running as fast as possible on CPU or GPU. It cleverly uses a series of code optimizations to squeeze as much performance out of the hardware as possible. If you're interested in the basic facts of mathematical optimization in code, check out this interesting list.
Theano is specifically designed for the computations required to handle large neural network algorithms in deep learning. It is one of the first libraries of its kind and is considered the industry standard for deep learning research and development.
Therefore, everyone calls Theano the grandfather of the deep learning framework!
A farewell letter, Theano disappears?
An email from Professor Yoshua Bengio announced Theano's historic mission. If you think Theano is dead, you are dead wrong!
From the current mainstream models, we can still see the shadow of Theano. It's not dead, but affects many, many models. In fact, many of Theano's developers went to Google to work on TensorFlow, including early developer Ian Goodfellow. The rising star Tensorflow is very similar in function to Theano, and its performance is more optimized.
From Theano to Tensorflow, a horizontal comparison of seven deep learning frameworks
Matt Rubashkin (Ph.D. UC Berkeley), a data engineer from Silicon Valley Data Science, a data science company, conducted a side-by-side comparison of 7 popular frameworks including Theano, TensorFlow, Torch, Caffe, MXNet, Neon, and CNTK.
The end result: If you want to get started with deep learning, you should start by assessing your own team skills and business needs. For example, if a Python-centric team wants to develop an application for image recognition, you should use TensorFlow because of its rich resources, better performance, and complete prototyping tools.
If you are using Tensorflow, then you must learn Theano!
main content
1. "Deep Learning Based on Theano: Building Future and Current Artificial Brains" introduces the basic concepts of deep learning and the application of Theano in deep learning.
2. "Deep Learning Based on TensorFlow: Revealing the Secrets of Data" introduces the introductory knowledge of TensorFlow and its application in deep neural networks, convolutional neural networks, and recurrent neural networks, and analyzes it in detail through specific examples with application.
About the Author
1. Christopher Bourez holds a Master's degree in Mathematics, Machine Learning and Computer Vision (MVA) from École Polytechnique and Ecole Normale Supérieure de Cachan. Under the active advocacy of Packt Press, the successful experience of the Caffe, TensorFlow or Torch tutorials written by it has been transplanted into this book on Theano technology.
2. Dan Van Boxel is a data analyst and machine learning engineer with over 10 years of development experience, most notably Dan Dose Data, a livestreaming platform on YouTube that demonstrates the power and pitfalls of neural networks.
编辑推荐
Theano是一个能够在CPU或GPU上便于优化数值表示和深度学习模型的Python库。《基于Theano的深度学习：构建未来与当前的人工大脑》提供了一些实用代码示例，有助于初学者易于理解如何构建复杂神经网络，而对于有经验的数据分析师会更关注书中的相关内容，解决图像识别、自然语言处理和博弈决策领域的监督式学习和非监督式学习、生成模型和强化学习。
阅读本书将会学到的内容：
•熟悉Theano和深度学习的概念；
•给出监督式、非监督式、生成或强化学习的示例；
•揭示设计高效深度学习网络的主要原则：卷积、残差连接和递归连接；
•Theano在实际计算机视觉数据集中的应用，如数字分类和图像分类；
•将Theano扩展到自然语言处理任务，如聊天机器人或机器翻译；
•人工智能驱动策略以使得机器人能够解决博弈问题或从环境中学习；
•基于生成模型生成真实的合成数据；
•熟悉应用于Theano上层的两个框架：Lasagne和Keras。
深度学习是目前的热点研究领域之一，而TensorFlow是由Google公司开发，研究深度学习的重要开源软件库。《基于Theano的深度学习：构建未来与当前的人工大脑》是在作者Dan的TensorFlow畅销视频课程基础上编著完成的。作者介绍了各种复杂的深度学习算法以及各种深度神经网络的应用案例，分享了其宝贵的经验和知识，通过实践示例的帮助下，你将成为在先进多层神经网络、图像识别以及其他方面的高手。
阅读本书将会学到的内容：
•配置计算环境和安装TensorFlow；
•构建日常计算的简单TensorFlow图；
•基于TensorFlow的逻辑回归分类应用；
•利用TensorFlow设计和训练多层神经网络；
•直观理解卷积神经网络在图像识别中的应用；
•神经网络从简单模型到更精准模型的改进；
•TensorFlow在其他类型神经网络中的应用；
•基于一种TensorFlow高级接口——SciKit Flow的神经网络编程。
本期荐书评奖规则
在本文下方留言，用30+个字符，描述你想要这本书的理由。
We will select 5 lucky winners ranked 3rd, 5th, 9th, 11th and 15th from the users who leave a message according to the number of likes on the message , and send 1 of the 2 recommendation books in this issue.
Prize draw time: May 4th
If you do not win the lottery, you can also click " Read the original text " below to buy at a 50% discount.
Half-off discount period: May 3 to 6 (if you want to buy it as soon as possible)
There is a reading exchange group here, would you like to know?
If you cannot scan the code to enter the group, you can add and edit WeChat, and remark #deep learning#. After the editor has verified it, you will be added to the group.
Program WeChat ID: druidlost
Xiaoqi WeChat ID: duoshangshuang