How to get started deep learning?

Tel-aviv University Deep Learning laboratory Ofir students wrote an article on how to get started in-depth study, translation, the benefit of biological information dog.

Artificial neural networks have recently made breakthroughs in many areas, such as facial recognition, object discovery, and go, and deep learning has become hot. If you are interested in deep learning, this article is a good starting point.

If you have studied linear algebra, calculus, Probability theory and programming , I suggest you start with the CS231N course at Stanford University. The course is extensive and highly written. Each lesson can be downloaded from the slide, although the official website deleted the matching video, but you can easily search the Internet.

If you haven't studied math classes , there are lots of free learning materials on the Web that you can use to learn the necessary math knowledge. Gilbert Stange's linear algebra course is perfect for getting started. For other subjects, EdX has an MIT calculus and probability theory course.

If you want to learn more about machine learning , Andrew Ng's machine learning courses on Coursera are the best choice for getting started. In addition, Yaser Abu-mostafa's machine learning program is more focused on theory, but also suitable for beginners. Learning deep learning does not require mastering machine learning, but it can be helpful if you have some understanding. In addition, Learning Classic machine learning, not just deep learning, can give us a deep theoretical background ———— because deep learning is not always the best solution.

cs231n is not the only option for online deep learning courses. Geoffrey Hinton's Coursera Course "neural Networks for Machine learning" (neural Nerworks for machines learning) covers a wide range of content, and Hugo Larochelle's "Neural Network Lesson" ( Neural Networks Class) is also true. Both courses have video materials. Nando de Freitas's online courses also come with videos, slideshows and homework.

If you don't like watching video, but prefer reading , neural networks and deep learning (neural Networks and Learning) is an online free book written for beginners in deep learning. The book of Deep Learning is also a great free book, but a little higher-learning.

Once you have the basics, you can also develop in these areas:

• Almost all of these deep learning materials are more or less related to computer vision (computer vision).
• Recurrent neural Networks (Recurrent nerual Networks) is the basis of a neural network model for solving problems such as machine translation and speech recognition. Andrej Karpathy's blog post on RNN can help you learn it. Christopher Olah's blog has an article explaining a lot of deep learning concepts in a very vivid way. His article on the LSTM network is a very good introduction, and Lstm is a widely used RNN variant.
• Natural Language Processing : The CS224D course introduces the application of deep learning in natural language processing. Higher-level courses come from Kyunghyun Cho (which has lecture notes) and Yoav Goldberg.
• Enhanced Learning : Enhanced learning may be the best option if you want to control a robot, or to defeat humans in a go game. Andrej Karpathy's blog post on Deep enhancement learning can help you get started. David Silver has also recently published a short article on deep-enhanced learning.

Deep Learning Framework : A lot of deep learning frameworks, the most famous three should be TensorFlow (Google), Torch (Facebook) and Theano (MILA). Three are very good, if must recommend one, I suggest beginners to try TensorFlow. TensorFlow's tutorial is very good.

Training a neural network is almost inseparable from the GPU. While not necessary, the GPU can help you get the job done faster. NVIDIA graphics cards are industry standards, and most research labs use a $1000 graphics card, and few bargains can fix it. Another way to lower costs is to rent an instance with a GPU (a short tutorial here) from a cloud service provider like Amazon.

Good luck!

June 26, 2016

Original address:/httpofir.io/how-to-start-learning-deep-learning/

Chengang
Links: https://zhuanlan.zhihu.com/p/21475898
Source: Know
Copyright belongs to the author. Commercial reprint please contact the author for authorization, non-commercial reprint please specify the source. Aa

How to get started deep learning?