Abstract
The given work describes a technology of construction of neural network system of artificial intellect (AI) at a junction of declarative programming and machine training on the basis of modelling of cortical columns. Evolutionary mechanisms, using available material and relatively simple phenomena, have created complex intelligent systems. From this, the authors conclude that AI should also be based on simple but scalable and biofeasible algorithms, in which the stochastic dynamics of cortical neural modules allow to find solutions to of complex problems quickly and efficiently.. Purpose: Algorithmic formalization at the level of replicative neural network complexes - neocortex columns of the brain. Methods: The basic AI module is presented as a specialization and formalization of the concept "Chinese room" introduced by John Earle. The results of experiments on forecasting binary sequences are presented. The computer simulation experiments have shown high efficiency in implementing the proposed algorithms. At the same time, instead of using for each task a carefully selected and adapted separate method with partially equivalent restatement of tasks, the standard unified approach and unified algorithm parameters were used. It is concluded that the results of the experiments show the possibility of effective applied solutions based on the proposed technology. Practical value: the presented technology allows creating self-learning and planning systems.
Highlights
Lucas A. et al Neural Networks for Modeling Neural Spiking in S1 Cortex // Front
Gradient-Based Learning Applied to Document Recognition // Proceedings of the IEEE. 1998. vol 86. no. 11. pp. 22782324
The basic artificial intellect (AI) module is presented as a specialization and formalization of the concept "Chinese room" introduced by John Earle
Summary
Что нейронная сеть с обратным распространением может решать сложные задачи нелинейного отображения, однако с увеличением размера выборки время, необходимое для обучения нейронных сетей, увеличивается, а точность классификации уменьшается. Что при моделировании нейронных систем на уровне бионейрона за деталями функционирования элементарных клеток могут теряться базовые принципы обучения, поэтому моделирование на уровне более крупных модулей может дать больше результатов. Записи в предпамять добавляются только при ненулевом количестве медиатора Good или Evil на входе нейромодуля. ; Output={,,...,}, где Input – необходимый состав информационных медиаторов на входе нейромодуля для активации правила; Output – список пар ; Mi – i-й вариант медиатора продукции; Ri – рейтинг i-го варианта медиатора продукции. На рисунке 3 иллюстрируется проблема, что если нейроны выстроены в длинную цепочку, то вероятность того, что изменения на входе цепочки приведут к изменениям на ее выходе, снижается по мере удлинения цепочки
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
