TensorFlow – the American giant Google’s web arrives in the field of machine learning (machine learning), namely the ability of computers to conceive and understand the information being able to process them in order to improve the performance with the passage of time. The new Google platform is called TensorFlow today Big G makes open source technology such artificial intelligence, a project which will now be able to access and therefore improve all developers in the world, from simple fan to the researcher university.
TensorFlow is definitely a tool that takes advantage of many very large algorithms based on single machine with different calculation options, such as Google Maps (offline browsing today), Google Now . capable of interacting with the user, planning, counseling, in short an assistant not only voice, but an agenda able to predict and anticipate our demands, still, Smart replay Gmail, able to “think” and therefore suggest the answers to be sent to the email we receive
“The machine learning is still in its infancy, can not do what a child of four years ago as easily recognize a dinosaur after seeing in the picture a couple of times. We have much work to do but TensorFlow is a good start and will help us to do this work together “, he explains in a post official Sundar Pichai , Google CEO. “Hopefully,” he adds, “that the community working on machine leraning exchange ideas quickly as possible to accelerate research. TensorFlow is a good platform not only for technology but is also useful for researchers who want to give meaning to their amount of data, the findings on the protein to those concerning astronomy. “
L ‘ machine learning (known in the literature as machine learning ) is one of the key areas of artificial intelligence and deals with the implementation of systems and algorithms that are based on observations as the data for the synthesis of new knowledge. Learning can take place by capturing characteristics of interest from examples, data structures or sensors, to analyze and assess the relationships between the observed variables.
General
One of the main objectives of research on machine learning is to learn to automatically recognize complex patterns and make intelligent decisions based on data; the difficulty lies in the fact that the set of all possible behaviors given all possible inputs is too large to be covered by sets of observed examples (training data). From here it is necessary the use of techniques to generalize the examples mentioned, so as to be able to produce a useful behavior for new cases.
Nowadays we are not yet able to reproduce systems Machine learning similar to the human. However they were invented algorithms effective for some types of learning tasks, so significant commercial applications have begun to appear.
For problems such as speech recognition, based on machine learning algorithms work best. In the area known as data mining, these algorithms are used routinely to discover valuable knowledge from large commercial databases containing a large amount of information.
The machine learning itself is a multidisciplinary field. It is based on the results of artificial intelligence, probability and statistics, computational complexity theory, control theory, information theory, philosophy, psychology, neurobiology, and other fields.
Examples of practical applications
Voice recognition text
All voice recognition systems most successful using machine learning methods. For example, the SPHINXsystem learn strategies specific speakers for recognizing primitive sounds (phonemes) and the words in the speech signal observed. Learning methods based on neural networks and hidden Markov models are effective for automatic customization of vocabulary, characteristics of the microphone, noise, etc.
Guide automatic vehicles
Machine learning methods were used to train computer-controlled vehicles. For example, the system ALVINN used his strategies learned to ride without assistance at 70 mph for 90 miles on public roads, among other cars. Similar techniques have potential applications in many control problems based on sensors.
Classification of new astronomical structures
Machine learning methods have been applied to a variety of large database to learn general regularities implicit in the data. For example, learning algorithms based on decision trees were used by NASA to classify celestial objects from the second Palomar Observatory Sky Survey. This system is now used to automatically classify all objects in Sky Survey, which consists of three terabytes of image data.
Player backgammon world class
The most successful computer programs for playing backgammon are based on learning algorithms. For example, the best computer program in the world for backgammon, TD-Gammon, developed its strategy by playing more than one million test matches against himself. Similar techniques have applications in many practical problems in which space research very important to be examined efficiently.
No comments:
Post a Comment