Yazar "Erkmen, Burcu" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • [ X ]
    Öğe
    A Stochastic Computing Method For Generating Activation Functions in Multilayer Feedforward Neural Networks
    (Istanbul Univ-Cerrahpasa, 2021) Ersoy, Durmuş; Erkmen, Burcu
    Stochastic computing using basic arithmetic logic elements based on stochastic bit sequences provides very beneficial solutions in terms of speed and hardware cost, relative to deterministic calculation. Studies for the realization of tangent hyperbolic and exponential functions used in the development of activation functions in Artificial Neural Networks by stochastic methods exist in the literature. The techniques presented using state transitions on finite state machines were constructed on the basis of two different forms of finite state machines, one-dimensional (Linear) and two-dimensional. In this analysis, in terms of both error rate and circuit cost, the advantageous two-dimensional finite state machines based stochastic computing approach for tangent hyperbolic and exponential functions is presented. The presented approach is implemented on Field Programmable Gate Array and the results are given for hardware simulation. The dataset used for the classification process in a decentralized smart grid control has been applied to the multilayer feedforward neural network and deterministic computing, for the stability classification which is carried out separately with the linear finite state machines based stochastic computing and the proposed 2D finite state machines based stochastic computing methods.
  • [ X ]
    Öğe
    Stochastic Gaussian Function for RBF Network
    (Institute of Electrical and Electronics Engineers Inc., 2020) Ersoy, Durmuş; Erkmen, Burcu
    In Artificial Neural Network applications, new solutions are searched for high speed and low circuit cost for high density inputs. In this study, a new Gaussian Function calculation method is presented for Radial Basis Function Network using stochastic calculation. The Gaussian Function of the Radial Basis Function Network was obtained using a Linear Finite State Machine approach. Stochastic representations of input values and centers were applied to XOR, OR and AND gates to realize simple arithmetic operations. The accuracy of the presented method depends on the bit length of the stochastic sequences. Using this method, considerable flexibility has been provided to the designer in terms of speed and hardware cost for applications with high input data. From the FPGA application results, the recommended stochastic calculation hardware resource requirement for the Gauss Function is much less than the hardware requirement of the corresponding deterministic calculation. The proposed stochastic network can be expanded to the large scale networks for complex tasks using simple hardware architectures. Simulation results and resource usage of FPGA are demonstrated in this paper. © 2020 IEEE.