Search

Search Results

• THE COMMON ALTERNATIVE APPROACH FOR THE EFFICIENT DEEP LEARNING NEURAL NETWORKS

  N.S. Krivsha, V.V. Krivsha, S.A. Butenkov

  2024-01-05

  Abstract ▼

  This paper proposes a new approach to the organization of computational structures of layers
  and inter-layer connections in the construction of artificial neural networks for solving a wide
  range of problems of multidimensional data processing. The main problem of building deep learning
  networks is the necessity of introducing a large number of network training parameters. Available
  working instances of such networks contain billions of parameters, which allows to achieve
  high efficiency of such networks. The downside of such a widely used structure of networks in the
  form of multilayer sieve structures is the high cost of training networks with a large number of
  structurally similar convolution layers by the back-propagation method. A solution to the problem
  of increasing the efficiency of such multilayer structures can be found in the use of hybrid layers
  realizing data granularity operations, which were developed in our work. The new hybrid models
  use matrix information elements instead of vector values of training parameters, which allow encoding
  subsets of data values (information granules) instead of encoding individual data points as
  in classical convolutional networks. The proposed hybrid layers are trained without a teacher and
  allow parallel implementation of learning algorithms, which is fundamentally different from sequential
  backpropagation algorithms as a result, the computational efficiency of similar hybrid
  neural networks can be significantly increased. The theoretical approach to modeling and optimizing
  the structures of deep learning networks proposed in this paper can be extended to a wide
  range of computational intelligence problems.

• THE STRUCTURE OF CUBATURE FORMULAS MODELLING FOR THE EFFICIENT FPGA IMPLEMENTATION

  N. S. Krivsha , V.V. Krivsha , S. A. Butenkov

  2021-01-19

  Abstract ▼

  In the paper we present the new computing models for the common cubature formulas computing
  unit design and optimization. The basis of new modeling technique is related with the space
  granulation theory, developed in our recent papers. The Spatial Granulation Technique allows us
  to pass from computing in the metrical data points space to affine data space, contains the aggregated
  data units named as granules. The introduced data transformation based on the affineinvariant
  Cartesian granule model and on the optimal data points coarsening procedures. The
  useful properties of new data models allows to provide the very efficient multivariable data management
  procedures. The one of them is the multivariate cubature formulas calculation. The new
  theory provides the obvious matrix data processing models for the information graphs design andoptimization. We can perform the equivalent mappings for the complicated information graph
  models for the efficient structures matching. Optimized models of information graphs are used for
  the FPGA-based devices implementation. The main problem of FPGA design is the commutation
  structures complication for the large FPGA fields, obtained as the basic units for the reconfigurable
  cubature formulas computing units. In this work we use the high-level programming language
  COLAMO and assembler language Fire Constructor for the computing units implementation. As a
  result of new technique implementation we can provide the family of adequate and useful graphic
  representation for a multivariable cubature formulas over the matrix calculation. The provided
  models are suitable for the optimal design of configurable computing structures, universal and
  dedicated devices from the FPGA basis. For the device implementation the developed high-level
  software products are used. For the designed universal devices the testing procedures was performed
  and examined with the symbolic calculation software for the computing results evaluation.