How to get a quick start in deep learning for people who are not good at math or English?

since the beginning of the 2016, the use of neural networks and deep learning Alphago to win the Master of Human go, deep learning is also considered to be the closest machine learning approach to AI.

from the current development trend of the global AI, the Deep learning framework is showing the trend of crowded, Google, Facebook, Microsoft, Amazon, Baidu and other large enterprises and a few start-up companies have launched their own products. Accordingly, the global name Enterprise to the Deep study field talented person also is lookout, is not difficult to see: profoundlearning not only is the new round artificial intelligence crosses the development core engine, is also the global science and technology innovation and the industry Development frontier position.

However, while the prospects for deep learning are very bright, many aspire to this field due to the high learning threshold Beginner are stuck in the entry period for several reasons:

1. Two basic obstacles: math and English

Deep LearningIt does require a certain mathematical basis. at least tounderstand scalar, vector, matrix, and tensor, familiar with the related operations (such as addition, multiplication, transpose, etc.), know what the feature decomposition and singular value decomposition are going to be and so on. But for the practicalDeep Learningthe mathematical part,ifThe study material is notin layman 'scome tospeaking,a lot of beginnersthere will be fear .sentimentso that it is easy to give up prematurely.

The more painful is that much about Deep learning lectures, books are from abroad, full screen of all English terminology is often to discourage beginners, and the rapid development of the deep learning field, learning from paper and recently published papers to understand the development of the forefront is also very necessary, If the English Foundation is not good, the best way is to find a big coffee to follow, learn what he has understood, chewed knowledge, for small white will be easy to digest much.

2, lost in the vast amount of information:

Deep Learning is a lot of information, but it's also a deep learningCompare Tai Hangthe place,Beginnercan easily get lost in all kinds of information. deep Learning is a discipline that combines theory with practice,It makes no sense to find a lot of books and resources, and it's easy to have a feeling of irritability when you look at several courses. soIt's best to cover all the materials you choose.theoryKnowledgeand Practicesection,or a book that tells the theory, and then find some code to do the experiment., of course, the latter kind of relatively difficult to some;

3. Learning Path Chaos:

many Deep learning beginners do not realize the importance of learning path, the wrong learning path will not only let you go a lot of long way, may even put your ideological pattern bound to meet the ceiling early.

the success of an area depends on its own efforts as well as on the course of history. the simplest and most effective path for deep learning is to learn from the process of history.

frombasic knowledge of machine learningStart , before learning and understanding deep learning,beginners should at least Learn about the two models of neural network and Autoencoder. Next Stepisunderstand what the problem is with "simply increasing the depth of the neural network." Like gradient dispersion and gradient explosions .,severe over-fittingwait. To understand the causes and solutions of these problems,to be Mapping to the development of neural networks to deep learning , such as the emergence of pre-training, dropout, relu, and various optimization algorithms such as Adam, Rmsprop, and so on, we now have the ability (computational power, Handle deep networks relative to previous large data volumesand so on.

Public Number: Maibanzhang

How to get a quick start in deep learning for people who are not good at math or English?