Model for faster and more efficient self-learning artificial intelligence presented 27.02.2017
Model for faster and more efficient self-learning artificial intelligence presented

Modern learning models for artificial intelligence are based on the analysis of big data. Algorithms process the data available online and try to forecast certain events afterwards.

More efficient learning techniques for AI systems have been offered by the American startup Gamalon.

Company representatives have developed a machine learning technology that allows using a fewer number of examples. Thus, if artificial intelligence is shown a picture if blue sky with white clouds, it will learn to identify sky on other similar images. Accumulating experience the algorithm will improve its knowledge and return more accurate data. In the process of learning the program rewrites the code on its own.

Gamalon’s developments are based on Bayesian program synthesis – a mathematical calculation of probability with due regard for already obtained experience. In other words, program uses probable values rather than defined values. This method was developed by a talented XVIII century’s mathematician – Thomas Bayes.

Gamalon created two products based on the innovative method.

The first Gamalon’s product is a drawing application. It predicts what user wants to draw by analyzing key details of an object. Thus, if user draws a square and a triangle on top of it, the application will define that it is a house. A similar development has been released by Google recently. But the IT giant’s product compares the image with sketches from the database. Unlike it, Gamalon’s application is more universal and makes more accurate predictions.

Second product of the startup allows recognizing the text. For instance, algorithm analyses the device specification and defines which brand released it, what the name of the model is, and what characteristics it has. No other program in the world features such functionality.

Gamalon’s method has a huge potential for automating machine learning technologies. The program makes its own necessary calculations, rewrites the code and adapts to the environment and current conditions.

Gamalon’s developments are already used in practice. Thus, Avaya company uses artificial intelligence to correct inaccuracies in the names and addresses stored in corporate databases. Gamalon’s algorithm allows doing it in a matter of minutes. And Gamalon Match application, developed based on this algorithm, helps to structure products and prices in the stores.

AI-industry news and cases for you