AI systems have learned how to create smart models for ML: a digest for beginners.
In November, members of the Google Brain research project published the results of an AutoML experiment. They managed to create a system that generates new AI models using the reinforcement learning method . The algorithm implemented in this way is already coping with a task better than solutions completely written by man.
In this article, we will talk about the features of the AutoML system, as well as a selection of books and machine learning courses that will help you become more familiar with the technologies of artificial intelligence.
/ Flickr / hackNY.org / CC
')
Google ’s AutoML technology was introduced in May as a system for automating the creation of machine learning models. Even then she could design small neural networks that worked on a par with neural networks developed by humans.
The Google method is based on two neural networks that are in constant contact, the controlling and the controlled (affiliated). The child system is trained on the basis of feedback from the supervisor, which assesses the effectiveness of passing tests. This process is repeated thousands of times until the desired result is achieved. In the Google experiment, the neural network was engaged in the recognition of objects in streaming video: people, cars, traffic lights, and so on.
Google's AutoML is not the first such system. However, the uniqueness of the latest development of the Brain project lies in the fact that it does not just modify the existing models, but selects and modifies them.
Development of machine learning systems from scratch is a resource - intensive task . AutoML is designed to accelerate the development of new models of MO. In addition, it reduces the threshold for entry into the industry for future generations of researchers.
Jeff Dean, head of Google Brain, said technology will help companies build AI systems, even if they lack extensive experience. He also called automated machine learning one of the most promising areas for research.
According to Gartner, by 2020, 2.3 million new jobs will be created in the field of artificial intelligence and machine learning. The development of development based on neural networks has made the industry competitive, and the largest technology companies are willing to pay high salaries.
Google and Facebook are creating special courses on machine learning techniques to help their employees grow and attract new ones. The development of courses, the preparation of methodologies and the writing of books are also carried out by large world universities.
Therefore, further, we have gathered useful materials and resources that will introduce you to machine learning technologies. The selection is based on the recommendations of Hacker News residents and participants of specialized threads on Quora, Reddit and Habrahabr.
It also includes our own materials on the topic.
/ Flickr / bradleypjohnson / CC
Applied Regression Analysis, by Norman Draper and others.
A book with the basics of regression analysis. There are many examples, exercises and tests for self-examination.
"Linear algebra and its applications", Gilbert Strang (Gilbert Strang)
Machine learning algorithms are based on the principles of linear algebra. And although this book was published in 1980, its value - an explanation of the practical application of the material - has been preserved to this day.
“Artificial Intelligence: A Modern Approach”, Peter Norvig (Peter Norvig)
This tutorial on working with AI systems is used in thousands of universities around the world. According to the residents of HN, Norvig's work is the best way to begin to get acquainted with this topic.
Hands-on Machine Learning, by Aurelien Geron
A good start for future ML specialists who are going to work with the TensorFlow library. The author is a former Google employee who was responsible for classifying decisions on videos on YouTube.
Deep Learning Book, Ian Goodfellow et al
The book is designed to help students and specialists in the field of machine learning. It is recommended by the Google Brain team. According to Ilona Mask, this work is “the only comprehensive book on this issue.”
“High Algorithm” by Pedro Domingos
The material is located at the junction of the philosophy and expert field of AI systems. Its author is the holder of many awards for achievements in the development of the direction of AI, including the award Data Science Innovation Award.
“Machine learning. Textbook ", Peter Flach (Peter Flach)
The author is a professor in the field of artificial intelligence systems at the University of Bristol. He immerses the reader into the practical component of machine learning from the first pages, gradually increasing complexity. The book has many examples with illustrations.
Neural Networks and Deep Learning, by Michael Nielsen
The book teaches the reader the principles of the development of neural networks with specific examples. Nielsen is convinced that this approach is better than simply listing a long list of concepts. It is recommended by the Google Brain team.
Andrew Una Courses in Machine Learning
Andrew Eun is closely associated with the success of the Google Brain project. He is considered one of the most authoritative experts in the field of machine learning (Andrew projects ). Students will gain knowledge of efficient ML algorithms and experience in their practical application.
Basics of machine learning
This complex from the University of Washington is another often recommended course for beginners. He quickly acquaints with the most popular methods of machine learning.
Applied deep learning course fast.ai
The goal of the project is to reduce the shortage of specialists in AI, as well as to create automated solutions, as is done in Google Brain. One of the course authors is Jeremy Howard, an entrepreneur, business strategist, developer, and educator. The second author is Rachel Thomas, PhD in mathematics. Forbes called her one of "20 incredible women promoting AI research."
Designing Artificial Intelligence for Games
On the development of artificial intelligence for the gaming industry. The articles consider important AI concepts and ways to optimize them for operation on modern multi-core processors.
Creative Applications of Deep Learning
The course covers the key components of deep learning. The focus is on the practical application of algorithms in application development.
Collection of lectures on convolutional neural networks for visual recognition
A series of videos on Stanford University’s YouTube-themed channel dedicated to machine vision technology. The collection of lectures pays special attention to the study of models for such tasks as the classification of images.
CS188 - Introduction to AI
Collection of free materials from the University of California at Berkeley. Among the speakers are Dan Clay (Dan Clay), an American scientist in the field of natural language processing and an assistant professor of computer science at Berkeley. There are homework.
Practical Deep Learning For Coders
A course of lectures based on fast.ai, Prepared by the aforementioned Jeremy Howard.
Practical Machine Learning Tutorial with Python
A series of practical tips on machine learning. With examples and code in Python.
3Blue1Brown
This YouTube channel can be a good start to dive into machine learning from scratch. The channel has 500 thousand subscribers, and the number of fans of the original lessons is growing.
Machine Learning Recipes with Josh Gordon
Machine learning tips from a Google developer who acts as a technical evangelist for TensorFlow in the corporation.
4 cloud security trends
Consider the global experience of protecting data in the cloud. Developing areas in this area include encryption, infrastructure monitoring, automation, and machine learning .
Big Data: great opportunity or big deception
In this material we find out what the industry sees in Big Data. We will also talk about the areas in which they are successfully applied and how they appeared.
Artificial Intelligence: What Scientists Think About It
We talk about the myths about the work of AI, which scientists regularly encounter.
How neural networks are now used: from research projects to entertainment services
An article about how machine learning is used today: on YouTube recommender algorithms, neural networks in physics and medicine.
True neural network implementation from scratch in the C # programming language
As the author himself says, the post is intended for those who are familiar with the mathematical principles of the operation of neural networks. For better assimilation of the material, he advises first to study additional articles ( this and this ).
Introduction to Neural Networks
Details of creating a neural network from scratch. With examples, graphs, schemes.
In this article, we will talk about the features of the AutoML system, as well as a selection of books and machine learning courses that will help you become more familiar with the technologies of artificial intelligence.
/ Flickr / hackNY.org / CC
')
Machine teaches another machine
Google ’s AutoML technology was introduced in May as a system for automating the creation of machine learning models. Even then she could design small neural networks that worked on a par with neural networks developed by humans.
The Google method is based on two neural networks that are in constant contact, the controlling and the controlled (affiliated). The child system is trained on the basis of feedback from the supervisor, which assesses the effectiveness of passing tests. This process is repeated thousands of times until the desired result is achieved. In the Google experiment, the neural network was engaged in the recognition of objects in streaming video: people, cars, traffic lights, and so on.
Google's AutoML is not the first such system. However, the uniqueness of the latest development of the Brain project lies in the fact that it does not just modify the existing models, but selects and modifies them.
Why do you need AutoML technology
Development of machine learning systems from scratch is a resource - intensive task . AutoML is designed to accelerate the development of new models of MO. In addition, it reduces the threshold for entry into the industry for future generations of researchers.
Jeff Dean, head of Google Brain, said technology will help companies build AI systems, even if they lack extensive experience. He also called automated machine learning one of the most promising areas for research.
According to Gartner, by 2020, 2.3 million new jobs will be created in the field of artificial intelligence and machine learning. The development of development based on neural networks has made the industry competitive, and the largest technology companies are willing to pay high salaries.
Google and Facebook are creating special courses on machine learning techniques to help their employees grow and attract new ones. The development of courses, the preparation of methodologies and the writing of books are also carried out by large world universities.
Therefore, further, we have gathered useful materials and resources that will introduce you to machine learning technologies. The selection is based on the recommendations of Hacker News residents and participants of specialized threads on Quora, Reddit and Habrahabr.
It also includes our own materials on the topic.
/ Flickr / bradleypjohnson / CC
Books
Applied Regression Analysis, by Norman Draper and others.
A book with the basics of regression analysis. There are many examples, exercises and tests for self-examination.
"Linear algebra and its applications", Gilbert Strang (Gilbert Strang)
Machine learning algorithms are based on the principles of linear algebra. And although this book was published in 1980, its value - an explanation of the practical application of the material - has been preserved to this day.
“Artificial Intelligence: A Modern Approach”, Peter Norvig (Peter Norvig)
This tutorial on working with AI systems is used in thousands of universities around the world. According to the residents of HN, Norvig's work is the best way to begin to get acquainted with this topic.
Hands-on Machine Learning, by Aurelien Geron
A good start for future ML specialists who are going to work with the TensorFlow library. The author is a former Google employee who was responsible for classifying decisions on videos on YouTube.
Deep Learning Book, Ian Goodfellow et al
The book is designed to help students and specialists in the field of machine learning. It is recommended by the Google Brain team. According to Ilona Mask, this work is “the only comprehensive book on this issue.”
“High Algorithm” by Pedro Domingos
The material is located at the junction of the philosophy and expert field of AI systems. Its author is the holder of many awards for achievements in the development of the direction of AI, including the award Data Science Innovation Award.
“Machine learning. Textbook ", Peter Flach (Peter Flach)
The author is a professor in the field of artificial intelligence systems at the University of Bristol. He immerses the reader into the practical component of machine learning from the first pages, gradually increasing complexity. The book has many examples with illustrations.
Neural Networks and Deep Learning, by Michael Nielsen
The book teaches the reader the principles of the development of neural networks with specific examples. Nielsen is convinced that this approach is better than simply listing a long list of concepts. It is recommended by the Google Brain team.
Courses and guides
Andrew Una Courses in Machine Learning
Andrew Eun is closely associated with the success of the Google Brain project. He is considered one of the most authoritative experts in the field of machine learning (Andrew projects ). Students will gain knowledge of efficient ML algorithms and experience in their practical application.
Basics of machine learning
This complex from the University of Washington is another often recommended course for beginners. He quickly acquaints with the most popular methods of machine learning.
Applied deep learning course fast.ai
The goal of the project is to reduce the shortage of specialists in AI, as well as to create automated solutions, as is done in Google Brain. One of the course authors is Jeremy Howard, an entrepreneur, business strategist, developer, and educator. The second author is Rachel Thomas, PhD in mathematics. Forbes called her one of "20 incredible women promoting AI research."
Designing Artificial Intelligence for Games
On the development of artificial intelligence for the gaming industry. The articles consider important AI concepts and ways to optimize them for operation on modern multi-core processors.
Creative Applications of Deep Learning
The course covers the key components of deep learning. The focus is on the practical application of algorithms in application development.
Videos and lectures
Collection of lectures on convolutional neural networks for visual recognition
A series of videos on Stanford University’s YouTube-themed channel dedicated to machine vision technology. The collection of lectures pays special attention to the study of models for such tasks as the classification of images.
CS188 - Introduction to AI
Collection of free materials from the University of California at Berkeley. Among the speakers are Dan Clay (Dan Clay), an American scientist in the field of natural language processing and an assistant professor of computer science at Berkeley. There are homework.
Practical Deep Learning For Coders
A course of lectures based on fast.ai, Prepared by the aforementioned Jeremy Howard.
Practical Machine Learning Tutorial with Python
A series of practical tips on machine learning. With examples and code in Python.
3Blue1Brown
This YouTube channel can be a good start to dive into machine learning from scratch. The channel has 500 thousand subscribers, and the number of fans of the original lessons is growing.
Machine Learning Recipes with Josh Gordon
Machine learning tips from a Google developer who acts as a technical evangelist for TensorFlow in the corporation.
Articles
4 cloud security trends
Consider the global experience of protecting data in the cloud. Developing areas in this area include encryption, infrastructure monitoring, automation, and machine learning .
Big Data: great opportunity or big deception
In this material we find out what the industry sees in Big Data. We will also talk about the areas in which they are successfully applied and how they appeared.
Artificial Intelligence: What Scientists Think About It
We talk about the myths about the work of AI, which scientists regularly encounter.
How neural networks are now used: from research projects to entertainment services
An article about how machine learning is used today: on YouTube recommender algorithms, neural networks in physics and medicine.
True neural network implementation from scratch in the C # programming language
As the author himself says, the post is intended for those who are familiar with the mathematical principles of the operation of neural networks. For better assimilation of the material, he advises first to study additional articles ( this and this ).
Introduction to Neural Networks
Details of creating a neural network from scratch. With examples, graphs, schemes.
Source: https://habr.com/ru/post/343762/
All Articles