https://news.cnblogs.com/n/586233
Welcome to PyTorch, farewell to Theano, 2017 deep learning framework development overview 2018-01-02 15:06
Facebook open source PyTorch, become a strong enemy of TensorFlow
Theano completes its historic mission and retire
The ONNX open platform was born, trying to promote an open development ecology
CNTK, Keras, MXNet all ushered in major updates
Microsoft and Amazon launch Gluon deep learning library, Intel launches reinforcement learning framework Coach
For mobile terminals, Caffe2 and TensorFlow Lite have been open sourced successively
Man-machine duel: artificial intelligence technology analysis of Ke Jie 's AlphaGo human-machine battle comment GF Jingong Research Report
Artificial intelligence definition : The activity dedicated to making machines intelligent, and intelligence is the ability of an entity to function visionarily and appropriately in its environment.
1. The main reasons for the rapid development of artificial intelligence are: 1. The breakthrough of algorithms, 2. The growth of data, 3. The development of hardware, and 4. The popularity of open source software.
2. Machine learning is an important branch of artificial intelligence, which mainly studies how to make computers improve performance through learning from experience and data.
3. The development of artificial intelligence in recent years is mainly the development of machine learning technology. Artificial intelligence technology led by deep learning is advancing by leaps and bounds.
In traditional machine learning fields such as speech recognition and image recognition , deep learning has achieved a leap in machine learning performance, and the hedge fund field has begun to compete for artificial intelligence talents.
Fourth, machine learning can be divided into three categories: supervised learning, unsupervised learning and reinforcement learning.
Supervised Learning : Given a set of data, we know what the correct output should look like. Through learning, the relationship between input data and output data is established.
Unsupervised learning : It means that there is no given output information in the data sample, and it is hoped to mine information from the data.
Reinforcement learning : We use the "trial - feedback - optimization" approach to improve the policy through the feedback of the surrounding environment for trial and error, and finally find a satisfactory solution.
5. Major breakthroughs in deep learning in the field of machine learning
Language recognition, image recognition, unmanned driving, intelligent recommendation system, drug activity prediction, natural language processing, face recognition payment, medical image recognition, online advertisement click rate prediction, etc.
CPU GPU TPU Comparison:
CPU: Central processing unit, general-purpose computing, complex structure, good at processing computing tasks with complex steps and complex data dependencies
GPU: Graphics processor, dedicated computing, unified structure, high computing speed, good at image processing, large data throughput, more suitable for the application of artificial intelligence big data
TPU: Dedicated chip designed for machine learning applications, performs a single job, machine learning models run faster on the chip, with greater throughput