"The Turing awards are like Nobel of computer science and come to recognize a field that, after a stage of stagnation, has resurfaced again. Neural networks They have had an important past and they have a promising future. " The professor of Information Systems and Management of the Pablo de Olavide University (Seville) highlights the recognition made by the Turing award (endowed with one million dollars) to researchers in artificial intelligence Geoffrey Hinton, Yann LeCun and Yoshua Bengiowin.
Neural networks, present in a multitude of usual devices, as voice aids or the security systems of vehicles, try to imitate the human brain and have gone from working with simple structures (monolayer) to doing it with complex systems (deep learning) to, according to the teacher and also director of the Data Science Lab, identify voices or distinguish images among many other applications.
Its use has been generalized in the work of artificial intelligence to perform complex classifications, predictions and models of mechanical learning. It is about imitating the most complex organ (the human brain), which is why it has been called bio-inspired model or artificial neurons.
These systems re-emerge in 2004 thanks to the award-winning researcher and professor at the University of Toronto (Canada), Geoffrey Hinton, who developed a concept that has been working for half a century and oriented to mechanical learning and recognition of complex elements such as speech or image. Hinton created a research community to which were also honored with the Turing Prize: Yann LeCun, from the University of New York, and Yoshua Bengio, of the Montreal (Canada).
The Turing awards, granted by the Association for Computing Machinery, the largest group of computer professionals, has awarded the million dollars of the prize to the three researchers. The first works for Google; the second, for Facebook; and the third for IBM and Microsoft. The three companies count on neural networks as fundamental tools in their recognition, classification and control programs.
Neural networks are based on increasingly complex mathematical systems that can learn from the analysis of amounts of information, says Salmerón, who is also a researcher and member of scientific entities such as Internet Society, Association of Computing Machinery, Association of Logic Programming or International Rough Sets Society. This ability has swept the limits by which this discipline suffered a stagnation.
Some of the fields of application are medicine or smart cars. Researchers from the Mechanical Engineering Department of the University of Malaga (UMA) have developed a vehicle safety system inspired by impulse neural networks (spiking neural networks), artificial models that act in a very similar way to biological ones, when processing information.
This model tries to imitate the brain and analyzes more conditions than those perceived at a glance or those included in current security programs. In this way, for example, the mechanical decision to brake is taken not only in the presence of an object, but also before the data of speed, engine power, temperature and humidity.