Deep learning is a developing field in the development of artificial intelligence that is rapidly becoming popular in computer science. As a subcategory of machine learning, it addresses issues such as the use of neural networks to optimize speech recognition, computer vision, natural language processing, etc. In recent years, deep learning has contributed to the solution of such tasks as perception of objects, machine translation and voice recognition, whereas these research topics for a long time reluctant to succumb to the AI. Neural networks
In information technology a neural network is a system of programs and data structures as close as possible to the work of the human brain. A neural network typically uses a large number of processors working in parallel, each of which has its sphere of knowledge and their own access to data in a local storage device.
As a rule, the neural network is initially "trained", that is, it serves a large amount of data and rules about their relationship (e.g. "father is older than father"). After that, the program indicates network how to behave in response to external signals (e.g., data entered by the user of the computer interacts with the network) or can initiate activity (in the framework of the access to the outside world).
Deep and machine learning
to understand what deep learning, it is important to first separate it from other disciplines in the field of artificial intelligence.
One of the branches of artificial intelligence is machine learning when a computer retrieves knowledge in a controlled process. As a rule, in this case, the human operator, allowing the machine to learn by hundreds or thousands of training examples, and manually correcting errors.
Although machine learning has gained a dominant position in the field of artificial intelligence, it still has disadvantages. First, it takes a very long time. Second, machine learning might not be the true measure of computer intelligence, so it uses human ingenuity and abstract concepts, enabling the machine to learn.
Unlike machine learning, deep learning in most cases is uncontrollable. So, it is necessary to create an extensive neural network that allows computers to self-learn and "think" without requiring direct human intervention.
Deep learning is not like a computer program, says psychologist and specialist in the field of artificial intelligence, Gary Marcus. As a rule, computer code is written in accordance with the very strict logical steps. "But in deep learning, we see something entirely different. There is no set of instructions that say: if this is true, then do this", the scientist said.
Instead of linear logic, deep learning is based on theories about how the human brain operates. The program consists of interwoven layers of interconnected nodes. She learns by changing combinations of connections between nodes after each new experience.
Deep learning has shown potential as the basis for software that is able to work on the emotions or events described in the text (even if not explicitly expressed), recognize objects in photos and make sophisticated predictions about the possible future human behavior. Game in-depth education
In 2011, Google launched a project to study the brain "Google Brainproject", within which was created a neural network with the introduction of its algorithms, deep learning. It is famous for its ability to recognize concepts at a high level.
Last year, Facebook was created a section on the study of artificial intelligence. With the help of deep learning solutions was created to recognize persons and objects in the 350 million photos and videos daily uploaded in this social network.
Other examples of deep learning in action – a service by voice recognition, like Google Now and Apple Siri. The future of
Deep learning – a very promising field, and it will make Autonomous cars and robotic servants are real. These machines will still be limited, but what they will be a force, only a few years ago was considered improbable, and their appearance among the people coming with unprecedented speed. The ability to analyze huge data sets and use deep learning in computer systems that can adapt to new experiences without having to depend on a human programmer, will lead to significant scientific discoveries. They occur in many areas from more effective drugs to new materials and robots with a great perception of the world.
according to Livescience