Pulsed neural networks based on delta-sigma modulation with GHA learning rule and their hardware implementation

نویسندگان

  • Yoshimitsu Murahashi
  • Hirohisa Hotta
  • Shinji Doki
  • Shigeru Okuma
چکیده

あらまし ニューラルネットワークの並列処理を生かした高速演算処理を実現するためには,ソフトウェアの 直列演算ではなく,すべてのニューロンをハードウェアに並列実装することが望ましい.筆者らはこれまでに, ディジタル回路のハードウェア実装に適した ∆Σ 変調に基づくパルスニューラルネットワークを提案した.提案 したニューラルネットワークは,∆Σ 変調された 1 ビットのパルス信号で値が表現されるため小さな回路規模で 実現可能であり,1 ビットでありながら精度の良い演算を実現できる.本論文では,提案するニューラルネット ワークのハードウェア実装法を提案する.そして,GHA学習則を組み込んだニューラルネットワークを,FPGA 上にハードウェア実装し,その動作検証を行う.また,提案するニューラルネットワークと CPU を FPGA 上 に実装し,単位時間,単位回路規模当りに処理できるビット数を評価値として比較を行った.提案手法は CPU 上のソフトウェア実現に比べて 200 倍程度の評価値が得られた. キーワード パルスニューラルネットワーク,ハードウェア実装,∆Σ 変調,GHA 学習則

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Decrease in Hardware Consumption and Quantization Noise of Digital Delta-Sigma Modulators and Implementation by VHDL

A new structure is presented for digital delta-sigma modulator (DDSM). Novel architecture decreases hardware consumption, output quantization noise and spurs in Comparison to previous architectures. In order to reduce the delay, power consumption and increase maximum working frequency, the pipelining technique and the carry skip adder are used. Simulation proposed architecture shows that the qu...

متن کامل

Performance Scrutiny of Two Control Schemes Based on DSM and HB in Active Power Filter

This paper presents a comparative analysis between two current control strategies, constant source power and generalized Fryze current, used in Active Power Filter (APF) applications having three different modulation methods. The Hysteresis Band (HB) and first-order Delta-Sigma Modulation (DSM) as well as the second-order DSM is applied. The power section of the active power filter is viewed as...

متن کامل

Hardware implementation of a MLP network with on-chip learning

In this paper we propose a method to implement in FPGA circuits, a feedforward neural network with on-chip delta rule learning algorithm. The method implies the building of a neural network by generic blocks designed in Mathworks’ Simulink environment. The main characteristics of this solution are on-chip learning algorithm implementation and high reconfiguration capability and operation under ...

متن کامل

Vlsi Delta-sigma Cellular Neural Network for Analog Random Vector Generation

We present a cellular neural network architecture for parallel analog random vector generation, including experimental results from an analog VLSI prototype with 64 channels. Nearest-neighbor coupling between cells produces parallel channels of uniformly distributed random analog values, with statistics that are truly uncorrelated across channels and over time. The cell for each random channel ...

متن کامل

Efficient Realization of Sigma-Delta (∑-Δ) Kalman Lowpass Filter in Hardware Using FPGA

Sigma-delta (Σ-Δ) modulation techniques have moved into mainstream applications in signal processing and have found many practical uses in areas such as high-resolution A/D, D/A conversions, voice communication, and software radio. Σ-Δ modulators produce a single, or few bits output resulting in hardware saving and thus making them suitable for implementation in very large scale integration (VL...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:
  • Systems and Computers in Japan

دوره 36  شماره 

صفحات  -

تاریخ انتشار 2005