Radu Dogaru

A modified RBF neural network for efficient current-mode VLSI implementation

A modified RBF neural network model is proposed allowing efficient VLSI implementation in both analog or digital technology. This model is based essentially on replacing the standard Gaussian basis function with a piece-wise linear one and on using a fast allocation unit learning algorithm for determination of unit centers. The modified RBF approximates optimally Gaussians for the whole range of parameters (radius and distance). The learning algorithm is fully on-line and easy to be implemented in VLSI using the proposed neural structures for on-line signal processing tasks. Applying the standard test problem of the chaotic time series prediction, the functional performances of different RBF networks were compared. Experimental results show that the proposed architecture outperforms the standard RBF networks, the main advantages being related with low hardware requirements and fast learning while the learning algorithm can be also efficient embedded in silicon. A suggestion for current-mode implementation is presented together with considerations regarding the computational requirements of the proposed model for digital implementations.

Simplicial RTD-based cellular nonlinear networks

Recently, a novel structure called the simplicial cellular neural network (CNN) has been introduced , which permits one to implement any Boolean/Gray-level function of any number of variables. This paper is devoted to explore novel circuit architectures for the implementation of the simplicial CNN based on resonant tunneling diodes. The final objective is to implement a fully programmable CNN in a hardware platform based on nanoelectronic devices.

Pyramidal cells: a novel class of adaptive coupling cells and their applications for cellular neural networks

A significant increase in the information processing abilities of CNNs demands powerful information processing at the cell level. In this paper, the defining formula, the main properties, and several applications of a novel coupling cell are presented. Since it is able to implement any Boolean function, its functionality expands on those of digital RAMs by adding new capabilities such as learning and interpolation. While it is able to embed all previously accumulated knowledge regarding useful binary information processing tasks performed by standard CNNs, the pyramidal universal cell provides a broader context for defining other useful processing tasks, including extended gray scale or color image processing as well. Examples of applications in image processing are provided in this paper. Implementation issues are also considered. Assuming some compromise between area and speed, a VLSI implementation of CNNs based on pyramidal cells offers a speedup of up to one million times when compared to corresponding software implementations.

Systematic Design for Emergence in Cellular Nonlinear Networks

Cellular nonlinear networks are naturally inspired computing architectures where complex dynamic behaviors may emerge as a result of the local or prescribed connectivity among simple cells. Functionally, much like in biology, each cell is defined by a few bits of information called a gene. Such systems may be used in signal processing applications (intelligent sensors) or may be used to model and understand natural systems. While many publications focus on the dynamics in cellular automata and various applications, less deal with an important problem, that of designing for emergence. Put in simple words: How to choose a cell such that a desired behavior will occur in the cellular system. This book proposes a systematic framework for measuring emergence and a systematic design method to locate computationally meaningful genes in a reasonable computing time. Programs and application examples are provided so that the reader may easily understand the new concepts and develop her own specific experiments. An accessible language recommends it to a large audience including specialists from various interdisciplinary fields who may benefit from a better understanding of emergence and its applications to their specific field.

Hybrid cellular automata as pseudo-random number generators with binary synchronization property

This paper introduces random number generators based on hybrid cellular automata (HCA). It is shown that by properly mixing CA cells from two categories (two different rules) an automaton generating near maximal length pseudo-random sequences can be easily obtained. The resulting system is similar to the well-known linear feedback shift register (LFSR) but in addition it posses the binary synchronization property that makes it a valuable ingredient for many applications.

Chaotic Scan: A Low Complexity Video Transmission System for Efficiently Sending Relevant Image Features

A novel image scanning and transmission system is proposed, where the traditional raster scan is replaced with a new one, called a chaotic scan. The result is a low complexity image transmission system with encryption and spread spectrum capabilities well suited for compressed sensing applications. Due to the uncorrelated nature of the consecutively scanned pixels, it allows a form of progressive compression and fast discovery of the relevant image features using only a small fraction of the transmitted pixels. The key ingredient of the proposed system is a chaotic counter addressing the sensor array, based on a cellular automaton exhibiting a pseudo-random chaotic behavior and binary synchronization property.

An efficient finite precision RBF-M neural network architecture using support vectors

This paper investigates the effects of using limited precision for efficient implementations of the RBF-M neural network. This architecture employs only simple arithmetic operators and is characterized by simple LMS training in an expanded feature space generated by RBF functions centered around support vectors selected via a simple algorithm. The classification performances of our low complexity, finite precision architecture are similar and even better to those obtained using the more complex SVM.

Physical modeling of RTD-based CNN cells

Resonant tunneling diodes (RTDs) have intriguing properties which make them a primary nanoelectronic device for both analog and digital applications. We present a physics-based model of the RTD and study the universal cell circuit for Boolean CNNs which is proposed in a companion paper (Dogaru et al., 2000). In this circuit, the negative differential resistance of the RTD is fully exploited. Spice simulations confirm that it is capable of realizing a large class of linearly not separable Boolean functions.

CNN GENES FOR ONE-DIMENSIONAL CELLULAR AUTOMATA: A MULTI-NESTED PIECEWISE-LINEAR APPROACH

This paper introduces a novel CNN cell which guarantees the implementation of any local rule on three variables defined by a Boolean truth table. Moreover, since the output of the cell is completely specified by a simple mathematical formula, it is possible to develop a systematic theory for locating those regions in the CNN genes parameter space where complex behaviors may occur. The output cell formula is a simple piecewise-linear function, and for the case of a one-dimensional CNN the entire set of 256 CNN genes associated with the corresponding local Boolean functions are listed in this paper.

HCA101: A chaotic map based on cellular automata with binary synchronization properties

This paper reviews the properties of a novel chaotic map based on hybrid cellular automata (HCA101-map) and compares them against the widely known logistic map with finite precision implementation, from the perspective of their use as PN-sequences in communication systems. It turns out that our HCA-map is superior in any respect to the logistic map and in addition posses the binary synchronization property allowing to dramatically simplify acquisition circuits needed to reconstruct the phase of the PN-sequence in the receiver.