Development of image-logical neural networks for image processing, object recognition and classification, navigation, control and monitoring of unmanned aerial vehicles
One of the main problems of traditional deep neural networks is the possibility of unforeseen, random errors. Our mathematical apparatus allows us to solve this problem by creating figurative-logical neural networks that model the imaginative representation and thinking of a person. This device is based on logical information processing and is encoded by Boolean functions. Such software will perform only logical operations without complex numerical calculations performed during the learning process. The advantage of figurative-logical neural networks is an increase in reliability in operation and the elimination of accidental errors and artifacts.
Artificial Intelligence And Digital Services
Artificial intelligence
about two years.
Sales locations around the world
Exclusion of the appearance of random errors and artifacts, Reliability of object recognition and decision-making, a 4-bit neural network, Amenable to encoding with Boolean functions that can be used in software for neuromorphic devices, Each neuron in the network performs logical operations that are similar to elementary reasoning, Noise reduction in images, Hardware implementation in the form of an Embedend system microprocessor
1. Simonov, N.A. "Concept for use in an intelligent system", "Micronica", 2020, vol. 49, No. 6, pp. 459-473. , 2. Simonov N.A. "Application of the spot model to solve inverse problems". Sensors, 2023, Volume 23, № 3: 1247. , 3. Simonov N.A., Rusalova M.N. "Representation of mental images using the model of spots in psychology". Natural Systems of the mind, 2023, Volume 1, pp. 4-23. , 4. Simonov N.A. “Development of a mathematical apparatus with a figurative representation of information for neuromorphic systems". Russian Microelectronics, 2023, Volume 52, Supplement. 1, pp. S159–S162.
N.A.
