Turn learn from the machine to the depth of learning data compilation
Depth learning to learn from the machine data compilation
In the past six months, the blogger has been engaged in self-learning artificial intelligence-related knowledge. As the fiery Artificial Intelligence last two years, we are able to find a lot of information from the Internet, including: MOOC, blog and other bloggers also spent a lot of time to find a large number of resources from "from entry to advanced" learning the road. Here, bloggers according to their own learning experience, the summary of the information used in the film Bowen. Because bloggers current research is primarily aimed at the image, so the process of material selection will focus on knowledge of the image area, but bloggers below the recommended profile also covers the content on text, voice processing, the reader can selective reading to their needs.
First, introductory information
(1) Andrew Ng "machine learning" (NetEase cloud classroom)
Learning portal of choice, a comprehensive course covers the basics of machine learning and rarely involve highly theoretical or proven, can help beginners to quickly master this program by learning classical machine learning algorithm, and access to comprehensive machine learning related technologies understanding. Wu used when teaching is matlab (octave), and also the corresponding training course (can be found in the network above). By supporting the complete beginner exercises to teach courses related algorithms, to complete this exercise can further deepen learners' understanding of the basics of machine learning by guiding programming practice step by step. Bloggers feel the course is ideal for beginners as the preferred course.
(2) Stanford University CS231 Convolutional Neural Networks for Visual Recognition Course
Li Feifei set up by the research group at Stanford University course on deep learning (mainly image aspects) of. Although the final aim of the course is to introduce the convolution neural networks, but most will be introduced before the course content in detail and interpretation of some of the basics of machine learning, such as: cross-entropy loss, stochastic gradient descent, backpropagation and various response function, etc. these contents are great supplemental teaching content Andrew Ng teacher, you need to read and grasp. Among them, especially the bloggers think good look gradient descent, understand the process of operation of the algorithm is very important that a blogger's blog may be able to help the reader slightly. Video content of the course seems to be found in the B station above, but also with the program supporting the corresponding programming exercises (with python programming language).
(2) Andrew Ng "convolution neural network" (Netease cloud classroom)
Readers need for image correlation algorithm research, and certainly can not bypass the convolutional neural network. Andrew Ng teacher provides a special curriculum for convolution neural networks in professional Netease micro-cloud in the classroom. The curriculum and teaching style "machine learning" course is very similar, very suitable not have much background knowledge of the reader convolution neural network learning. In addition, the second half of the second courses also have specific content convolution neural network can be used with the contents of this course together.
Second, the practice of information
After learning of the above, I believe there is a preliminary master the basics of the readers of the depth of learning, then you can through some independent programming practices to internalize the knowledge acquired, makes them his own near life skills. Then bloggers on the introduction of two books and part pin Tensorflow Keras applications.
（1）Learning Tensorflow: A Guide to Building Deep Learning System(by Tom Hope, Yehezkel S.Resheff & Itay Lieder)
Tensorflow is the most widely used of deep learning framework, a lot of deep learning algorithms are applied to the frame, and thus one is engaged in the framework of relevant staff must master. Bloggers recommended this book seems to be Google's R & D personnel edited, the contents of the book introduces the installation Tensorflow framework, the basic components of the framework and how to apply at different depths learning tasks. The contents of each part of the book with a very detailed code for the reader, is very conducive to the practical operation of the learner, but the book also basic knowledge of some models were reviewed, the content can well help learners the basic concept combined with the practical operation.
（2）Deep Learning with Python(by Francois Chollet)
This framework presented in this book is Keras, which is built on a deep learning Tensorflow senior library (now a Tensorflow official high-level interface), compared to Tensorflow, the application of the modeling framework is more simple, intuitive, and It provides a wealth of functions for developers to use. Bloggers recommended by the author of the first edition of this book is written Keras library, the contents of the book covers all the content from machine learning to deep learning applications, and provides a lot of code examples in this book bloggers seems both a comprehensive review of machine learning to learning materials development context and depth of the basics, reference Guide is an application Keras library.
Third, the theory advanced
Compared to the recommended content "know and not know why," the characteristics of this part of the recommended data will focus on machine learning and deep learning more theoretical side, bloggers recommend the two main materials : "The statistical learning methods" and online book "Deep learning"
(1) "statistical learning methods" Lee Hang the
Li Hang teacher of this book focuses on machine learning theory analysis, the basic elements of machine learning, a common algorithm, a mathematical definition, analysis and deduction, through the study of these elements, I think let the reader these algorithms have a more profound understanding. Bloggers from their current reading experience point of view, the feeling is very difficult and requires a more solid mathematical skills to be able to understand the content of the teacher in the book to explain.
（2）《Deep Learning》（by Ian Goodfellow and Yoshua Bengio and Aaron Courville）
This book is a line, taking into account the depth of breadth and depth of learning content covers a very rich learning content. The first part of the general explained the basics necessary for the learner, including: linear algebra, probability theory and information theory, the reader can leak filled in accordance with the guidelines of this part. The second part describes the depth is very mature technology learning, comprising: a neural network, a convolutional neural network, a signal processing sequence. The third part describes the current depth study is still in the field of active research, including the characteristics of a learning and so on. Many contents of the book has a very detailed theoretical and mathematical derivation, bloggers feel is a very hard nut to book.
Finally, we offer bloggers to link two sites: Link 1 and Link 2 , you can find a wealth of free learning materials in them. Finally, I wish you all a happy learning and becoming the field Daniel!
Depth learning to learn from the machine data compilation
In the past six months, the blogger has been engaged in self-learning artificial intelligence-related knowledge. As the fiery Artificial Intelligence last two years, we are able to find a lot of information from the Internet, including: MOOC, blog and other bloggers also spent a lot of time to find a large number of resources from "from entry to advanced" learning the road. Here, bloggers according to their own learning experience, the summary of the information used in the film Bowen. Because bloggers current research is primarily aimed at the image, so the process of material selection will focus on knowledge of the image area, but bloggers below the recommended profile also covers the content on text, voice processing, the reader can selective reading to their needs.
First, introductory information
(1) Andrew Ng "machine learning" (NetEase cloud classroom)
Learning portal of choice, a comprehensive course covers the basics of machine learning and rarely involve highly theoretical or proven, can help beginners to quickly master this program by learning classical machine learning algorithm, and access to comprehensive machine learning related technologies understanding. Wu used when teaching is matlab (octave), and also the corresponding training course (can be found in the network above). By supporting the complete beginner exercises to teach courses related algorithms, to complete this exercise can further deepen learners' understanding of the basics of machine learning by guiding programming practice step by step. Bloggers feel the course is ideal for beginners as the preferred course.
(2) Stanford University CS231 Convolutional Neural Networks for Visual Recognition Course
Li Feifei set up by the research group at Stanford University course on deep learning (mainly image aspects) of. Although the final aim of the course is to introduce the convolution neural networks, but most will be introduced before the course content in detail and interpretation of some of the basics of machine learning, such as: cross-entropy loss, stochastic gradient descent, backpropagation and various response function, etc. these contents are great supplemental teaching content Andrew Ng teacher, you need to read and grasp. Among them, especially the bloggers think good look gradient descent, understand the process of operation of the algorithm is very important that a blogger's blog may be able to help the reader slightly. Video content of the course seems to be found in the B station above, but also with the program supporting the corresponding programming exercises (with python programming language).
(2) Andrew Ng "convolution neural network" (Netease cloud classroom)
Readers need for image correlation algorithm research, and certainly can not bypass the convolutional neural network. Andrew Ng teacher provides a special curriculum for convolution neural networks in professional Netease micro-cloud in the classroom. The curriculum and teaching style "machine learning" course is very similar, very suitable not have much background knowledge of the reader convolution neural network learning. In addition, the second half of the second courses also have specific content convolution neural network can be used with the contents of this course together.
Second, the practice of information
After learning of the above, I believe there is a preliminary master the basics of the readers of the depth of learning, then you can through some independent programming practices to internalize the knowledge acquired, makes them his own near life skills. Then bloggers on the introduction of two books and part pin Tensorflow Keras applications.
（1）Learning Tensorflow: A Guide to Building Deep Learning System(by Tom Hope, Yehezkel S.Resheff & Itay Lieder)
Tensorflow is the most widely used of deep learning framework, a lot of deep learning algorithms are applied to the frame, and thus one is engaged in the framework of relevant staff must master. Bloggers recommended this book seems to be Google's R & D personnel edited, the contents of the book introduces the installation Tensorflow framework, the basic components of the framework and how to apply at different depths learning tasks. The contents of each part of the book with a very detailed code for the reader, is very conducive to the practical operation of the learner, but the book also basic knowledge of some models were reviewed, the content can well help learners the basic concept combined with the practical operation.
（2）Deep Learning with Python(by Francois Chollet)
This framework presented in this book is Keras, which is built on a deep learning Tensorflow senior library (now a Tensorflow official high-level interface), compared to Tensorflow, the application of the modeling framework is more simple, intuitive, and It provides a wealth of functions for developers to use. Bloggers recommended by the author of the first edition of this book is written Keras library, the contents of the book covers all the content from machine learning to deep learning applications, and provides a lot of code examples in this book bloggers seems both a comprehensive review of machine learning to learning materials development context and depth of the basics, reference Guide is an application Keras library.
Third, the theory advanced
Compared to the recommended content "know and not know why," the characteristics of this part of the recommended data will focus on machine learning and deep learning more theoretical side, bloggers recommend the two main materials : "The statistical learning methods" and online book "Deep learning"
(1) "statistical learning methods" Lee Hang the
Li Hang teacher of this book focuses on machine learning theory analysis, the basic elements of machine learning, a common algorithm, a mathematical definition, analysis and deduction, through the study of these elements, I think let the reader these algorithms have a more profound understanding. Bloggers from their current reading experience point of view, the feeling is very difficult and requires a more solid mathematical skills to be able to understand the content of the teacher in the book to explain.
（2）《Deep Learning》（by Ian Goodfellow and Yoshua Bengio and Aaron Courville）
This book is a line, taking into account the depth of breadth and depth of learning content covers a very rich learning content. The first part of the general explained the basics necessary for the learner, including: linear algebra, probability theory and information theory, the reader can leak filled in accordance with the guidelines of this part. The second part describes the depth is very mature technology learning, comprising: a neural network, a convolutional neural network, a signal processing sequence. The third part describes the current depth study is still in the field of active research, including the characteristics of a learning and so on. Many contents of the book has a very detailed theoretical and mathematical derivation, bloggers feel is a very hard nut to book.
Finally, we offer bloggers to link two sites: Link 1 and Link 2 , you can find a wealth of free learning materials in them. Finally, I wish you all a happy learning and becoming the field Daniel!