Modeling of New Technical Systems Using Cause-Effect Relationships
Svetlana V. Davydova, Dmitriy M. Korobkin, Sergey Alekseevich Fomenkov, Sergey Kolesnikov · 2018
A unit of spatiotemporal associative memory is studied in the paper. The heart of the unit is a very simple and obvious dynamical system which may be considered as a grid with moving objects. As a result of objects' movement, preferable trajectories are formed, which are a basis of temporal associations (cause-effect relations). Developing connections between grid nodes and interface neurons determine spatial associations (similarity associations). It is shown that the unit is easily trained and retrained with its facilities approaching those for a more complex dynamic-neurons-based system. Also describes the process of documents filtering based on the SOM algorithm. It is known that self-organizing map - SOM is the neural network with unsupervised learning that performs the task of visualization and clustering. The idea of the network proposed by the Finnish scientist T. Kohonen. Is a method of projection of the multidimensional space into space with lower dimension (usually two-dimensional), it is also used for the decision of tasks of modeling, forecasting, etc., is a version of Kohonen neural networks. Self-organizing maps are used to solve tasks such as modeling, forecasting, identifying sets of independent attributes, data compression, and for finding patterns in large data sets. The most commonly described algorithm is applied for clustering data.