Wyszukiwarka
Filtry
wszystkich: 67
Wyniki wyszukiwania dla: CONVOLUTION
-
Geometric analogue of holographic reduced representation
PublikacjaHolographic reduced representations (HRRs) are distributed representations of cognitive structuresbased on superpositions of convolution-bound n-tuples. Restricting HRRs to n-tuples consisting of 1,one reinterprets the variable binding as a representation of the additive group of binary n-tupleswith addition modulo 2. Since convolutions are not defined for vectors, the HRRs cannot be directlyassociated with geometric structures....
-
Applications of the discrete green's function in the finite-difference time-domain method
PublikacjaIn this paper, applications of the discrete Green's function (DGF) in the three-dimensional (3-D) finite-difference time-domain (FDTD) method are presented. The FDTD method on disjoint domains was developed employing DGF to couple the subdomains as well as to compute the electromagnetic field outside these subdomains. Hence, source and scatterer are simulated in separate subdomains and updating of vacuum cells, being of little...
-
Acceleration of the Discrete Green’s Function Formulation of the FDTD Method Based on Recurrence Schemes
PublikacjaIn this paper, we investigate an acceleration of the discrete Green's function (DGF) formulation of the FDTD method (DGF-FDTD) with the use of recurrence schemes. The DGF-FDTD method allows one to compute FDTD solutions as a convolution of the excitation with the DGF kernel. Hence, it does not require to execute a leapfrog time-stepping scheme in a whole computational domain for this purpose. Until recently, the DGF generation...
-
Deep Learning Optimization for Edge Devices: Analysis of Training Quantization Parameters
PublikacjaThis paper focuses on convolution neural network quantization problem. The quantization has a distinct stage of data conversion from floating-point into integer-point numbers. In general, the process of quantization is associated with the reduction of the matrix dimension via limited precision of the numbers. However, the training and inference stages of deep learning neural network are limited by the space of the memory and a...
-
Digital structures for high-speed signal processing
PublikacjaThe work covers several issues of realization of digital structures for pipelined processing of real and complex signals with the use of binary arithmetic and residue arithmetic. Basic rules of performing operations in residue arithmetic are presented along with selected residue number systems for processing of complex signals and computation of convolution. Subsequently, methods of conversion of numbers from weighted systems to...
-
FPGA implementation of the multiplication operation in multiple-precision arithmetic
PublikacjaAlthough standard 32/64-bit arithmetic is sufficient to solve most of the scientific-computing problems, there are still problems that require higher numerical precision. Multiple-precision arithmetic (MPA) libraries are software tools for emulation of computations in a user-defined precision. However, availability of a reconfigurable cards based on field-programmable gate arrays (FPGAs) in computing systems allows one to implement...
-
Cascade Object Detection and Remote Sensing Object Detection Method Based on Trainable Activation Function
PublikacjaObject detection is an important process in surveillance system to locate objects and it is considered as major application in computer vision. The Convolution Neural Network (CNN) based models have been developed by many researchers for object detection to achieve higher performance. However, existing models have some limitations such as overfitting problem and lower efficiency in small object detection. Object detection in remote...