Hardware-aware model of sigma-delta cellular neural network

Complex behavior of polynomial FitzHugh–Nagumo cellular neural network model

A model of cellular neural networks with mixed delays is proposed. Applying contraction mapping fixed point theorem and differential inequality technique, several sufficient conditions are derived to ensure the existence and exponential stability of weighted pseudo anti‐periodic solutions for the proposed neural networks. Finally, a numerical simulation example is given to illustrate the effectiveness of the obtained results.

The paper considers the full-range (FR) model of cellular neural networks (CNNs) with ideal hard-limiter non-linearities that limit the allowable range of the neuron state variables. It is also supposed that there is a concentrated delay (D) in the neuron interconnections. Due to the presence of multivalued nonlinearities the D-FRCNN model is mathematically described by a retarded differential inclusion. The main result is a rigorous proof that, in the case of nonsymmetric cooperative (nonnegative) interconnections, and delayed interconnections, the semiflow generated by D-FRCNNs is monotone, and that monotonicity implies some basic restrictions on the long-term behavior of the solutions. The result is compared with recent results in the literature on semiflows generated by cooperative standard CNNs, with and without delays.

Theoretic design of differential minimax controllers for stochastic cellular neural networks

In this paper we show cellular neural network (CNN) models of some basic types of cells characterized by diverse spiking patterns. After presenting some preliminary models (ion channels, neurons), CNN models of action potential generation are given, followed by an analysis of the rate-coding capabilities of the models. Furthermore, we show CNN models of synaptic connections. Finally, CNN models of neurons with diverse intrinsic firing patterns are presented.

We describe a novel cellular connectionist neural network model for the implementation of clustering-based Bayesian image segmentation with Gibbs random-field spatial constraints. The success of this algorithm is largely due to the neighborhood constraints modeled by the Gibbs random field. However, the iterative enforcement of the neighborhood constraints involved in the Bayesian estimation would generally require tremendous computational power. Such computational requirement hinders the real-time application of the Bayesian image segmentation algorithms. The cellular connectionist model proposed aims at implementing the Bayesian image segmentation with real-time processing potentials. With a cellular neural network architecture mapped onto the image spatial domain, the powerful Gibbs spatial constraints are realized through the interactions among neurons connected through their spatial cellular layout. This network model is structurally similar to the conventional cellular network. However, in this new cellular model, the processing elements designed within the connectionist network are functionally more versatile to meet the challenging needs of Bayesian image segmentation based on the Gibbs random field. We prove that this cellular neural network does converge to the desired steady state with a properly designed update scheme. Examples of CT volumetric medical image segmentation are presented to demonstrate the potential of this cellular neural network for a specific image segmentation application.

An optoelectronic model of discrete time cellular neural networks (DTCNN) is presented. Connections between cells and parallel input-output are realized using optoelectronic devices. As an emitter-receiver device an optical thyristor is applied. Connections between cells are realized by diffractive Damman gratings. We propose a dual rail system for early processing of binary images. The CNN system performs mathematical morphology and space logic operations.

The paper presents a method for increasing the resolution of parallel analog-to-digital converters (ADCs) with time compression of signals. It is established that analog-to-digital conversion of signals in parallel ADCs with time compression of output signals is performed in separate channels with non-identical bandwidths. Ultimately, this leads to a decrease in resolution when converting input signals. The difference in the bandwidths of parallel ADC channels is due to the features of the sample and store circuits (SSCs), which are installed at the inputs of each ADC. SSCs are necessary to reduce dynamic errors that arise when discretizing time-varying analog signals. The functioning of SSCs is based on the principle of fixing the instantaneous value of the input signal for the time required for further conversion in the ADC. The input circuits of SSCs can be represented as low-pass filters, in which there is a spread of the parameters of resistors and capacitors that determine the bandwidth of the i-th ADC channel. The passband adjustment in each ADC channel is carried out by switching resistors in the input circuits of the converter. To reduce the errors of non-identity of the passbands, it is proposed to introduce a reference additional ADC (АADC), which operates with a lower sampling frequency. In the operating mode, one of the ADC channels is calibrated. At the same time, the reference АADC operates with a sampling frequency that allows it to operate in phase with one of the working ADCs in each conversion cycle. A structural diagram of a parallel ADC (PADC) based on time compression of signals with correction of the passbands of individual ADC channels has been developed. The PADC operates in two modes: calibration and working with correction. In the calibration mode, a sinusoidal signal is supplied to the inputs of all PADC channels from the output of the test signal generator through an analog switch. The array of digital data from the output of the ADC is accumulated in a buffer storage device. Then, the accumulated array is fed to a computer via an interface adapter and a USB bus, where the errors of non-identity of the bandwidths of the ADC channels are calculated. Each ADC channel contains a digital-to-analog converter at the input, which performs the function of a digital potentiometer for the input signals of the SSC. The calculated transfer coefficients from the outputs of the control unit are fed to the corresponding ADC channels. Analysis of the ADC efficiency showed that by adjusting the bandwidths of individual channels of a 12-bit PADC, it is possible to improve the resolution in the low-frequency band by 0,5 bits, and in the high-frequency band by 1,9 bits.

An accelerated-time simulation of car traffic on a motorway using a CAS

A method for implementing Gaussian low-pass filter by using switched-current technique has been presented in this paper. The stable rational transfer function for Gaussian-shaped low-pass filter is approximated by the multiplication of the follower-integrator functions based on central limit theorem. An SI filter structure constructed from the inverse follow-the-leader feedback configuration has been proposed to synthesize the obtained transfer function, in which the switched-current bilinear integrator is employed as the building cell. Explicit design formulas are also derived. The 8th-order Gaussian low-pass filter is selected as the example to elaborate the design details. Simulation result shows that the proposed design method has high approximation precision and low magnitude sensitivity.

In the present work an adaptation of the cellular neural network (CNN) model to grey scale image processing is proposed. This task is performed programming the network to work as a classical spatial filter, taking advantage of the neural network structure in order to improve the filtering effects. This enhancement is carried out by the inclusion of the feedback of the state variables and the adaptation of the input bias of every neuron based on the brightness of the image. A proper choice of the gain of the output function may also improve some of the network capabilities.

Understanding urban expansion patterns is critical for sustainable land-use planning, particularly in rapidly developing regions. This study employs an integrated geographic ınformation system-based approach utilizing the MOLUSCE plugin in QGIS to predict future Land Use and Land Cover (LULC) changes in Sakarya Province, Türkiye, from 2011 to 2040. While cellular automata-artificial neural network and Markov Chain models have been extensively utilized in LULC research, their application to mid-sized Turkish provinces remains insufficiently explored. This study addresses that gap by providing the first comprehensive spatiotemporal analysis of Sakarya’s urban growth trends, incorporating environmental impact assessments alongside predictive modeling. The results indicate a 3.43% decline in agricultural areas and a 0.43% decrease in forests from 2011 to 2024, with continued urban expansion projected through 2040. These findings underscore the urgent need for sustainable zoning policies and conservation strategies to mitigate the adverse effects of urbanization on natural resources. The study not only enhances regional LULC modeling efforts but also provides a scalable framework for other developing urban centers facing similar land-use challenges.

The ADCs characteristics that make up the MC don't always correspond to the specified requirements for accuracy and noise level. The thesis paper describes a new type of ADC with phase-locked loop (ADC with PLL). Currently, the development of a number of electronic devices is conducted jointly with the use of various microcontrollers (MK), which includes analog-digital converters (ADCs). A new type of ADC allows improve these characteristics, having a relatively simple circuit design implementation. Using the PLL loop, the input level voltage (analog signal) is converted into a digital frequency signal of a variable pulse ratio. The converted signal (analog voltage) is applied to one of the low-pass filter (LPF) inputs included in the PLL, thereby removing the circuit from equilibrium, changing the pulse ratio of the output signal.

Over the past few decades, convolutional neural networks (CNNs) have achieved outstanding results in addressing a broad scope of computer vision problems. Despite these improvements, fully convolutional Siamese neural networks (FCSNN) still hardly adapt to complex scenes, such as appearance change, scale change, similar objects interference, etc. The present study focuses on an enhanced FCSNN based on convolutional block hyperparameters optimization, a new activation function (ModReLU) and Gaussian low pass filter. The optimization of hyperparameters is an important task, as it has a crucial ascendancy on the tracking process performance, especially when it comes to the initialization of weights and bias. They have to work efficiently with the following activation function layer. Inadequate initialization can result in vanishing or exploding gradients. In the first method, we propose an optimization strategy for initializing weights and bias in the convolutional block to ameliorate the learning of features so that each neuron learns as much as possible. Next, the activation function normalizes the output. We implement the convolutional block hyperparameters optimization by setting the convolutional weights initialization to constant, the bias initialization to zero and the Leaky ReLU activation function at the output. In the second method, we propose a new activation, ModReLU, in the activation layer of CNN. Additionally, we also introduce a Gaussian low pass filter to minimize image noise and improve the structures of images at distinct scales. Moreover, we add a pixel-domain-based color adjustment implementation to enhance the capacity of the proposed strategies. The proposed implementations handle better rotation, moving, occlusion and appearance change problems and improve tracking speed. Our experimental results clearly show a significant improvement in the overall performance compared to the original SiamFC tracker. The first proposed technique of this work surpasses the original fully convolutional Siamese networks (SiamFC) on the VOT 2016 dataset with an increase of 15.42% in precision, 16.79% in AUPC and 15.93% in IOU compared to the original SiamFC. Our second proposed technique also reveals remarkable advances over the original SiamFC with 18.07% precision increment, 17.01% AUPC improvement and an increase of 15.87% in IOU. We evaluate our methods on the Visual Object Tracking (VOT) Challenge 2016 dataset, and they both outperform the original SiamFC tracker performance and many other top performers.

CNNEDGEPOT: CNN based edge detection of 2D near surface potential field data