-
Research Article
Enhancing GAN Training Stability for Image Super-Resolution Reconstruction with AdaBelief Optimization Strategy
Issue:
Volume 11, Issue 2, December 2025
Pages:
42-50
Received:
10 August 2025
Accepted:
18 August 2025
Published:
23 September 2025
Abstract: Generative Adversarial Networks (GANs) have achieved remarkable success in image super-resolution reconstruction, producing high-quality images from low-resolution inputs. However, their training process is often plagued by instability issues, such as mode collapse and slow convergence, which hinder consistent performance. To address these challenges, we propose integrating the AdaBelief optimization strategy into GAN training to enhance both stability and the quality of generated high-resolution images. Unlike traditional optimizers, AdaBelief dynamically adjusts the learning rate based on the belief in observed gradients, enabling more precise and adaptive parameter updates for both the generator and discriminator. This approach mitigates the oscillatory behavior commonly observed during GAN training and improves the convergence properties of the adversarial learning process. We evaluated the proposed method on benchmark datasets, where it demonstrated superior performance in both quantitative metrics, such as peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM), and visual quality compared to conventional optimization techniques. Our experiments further reveal that AdaBelief fosters a more balanced rivalry between the generator and discriminator, promoting stable training dynamics and reducing the risk of mode collapse. This work is significant for advancing the practical application of GANs in super-resolution tasks, where stable training and high-fidelity outputs are critical. By offering a robust and efficient alternative to existing optimizers, AdaBelief addresses persistent challenges in GAN training, paving the way for more reliable and effective image super-resolution solutions. Our findings underscore the potential of AdaBelief as a versatile optimization strategy, delivering consistent improvements across diverse datasets and applications.
Abstract: Generative Adversarial Networks (GANs) have achieved remarkable success in image super-resolution reconstruction, producing high-quality images from low-resolution inputs. However, their training process is often plagued by instability issues, such as mode collapse and slow convergence, which hinder consistent performance. To address these challeng...
Show More
-
Research Article
Formation and Manifestations of Natural Human Intelligence Under the Influence of the Universe
Evgeny Bryndin*
Issue:
Volume 11, Issue 2, December 2025
Pages:
51-57
Received:
27 August 2025
Accepted:
9 September 2025
Published:
9 October 2025
Abstract: Intelligence, consciousness, knowledge and skills are key concepts related to understanding human activity and cognitive processes. Consciousness is a subjective perception and awareness of the surrounding world, one's sensations, thoughts and feelings. It allows a person to feel like a separate person, to be aware of their actions and experiences. Intelligence is the ability to analyze, think logically, solve problems, adapt to new conditions and be creative. Intelligence can use knowledge, draw conclusions and make decisions. Knowledge is a set of information, facts, skills and abilities acquired by a person through experience, training or inheritance. Knowledge serves as a basis for thinking and acting. There is a connection between them. Consciousness provides awareness and perception of knowledge. Intelligence uses knowledge for analysis, problem solving and creative activity. Knowledge is replenished and updated through consciousness, thinking and the activity of the intellect. Thinking is the process of processing information that involves using intelligence and knowledge to form new ideas, solutions, and concepts. Together, these concepts form the basis for understanding the human mind and its functioning. Human consciousness is firmly connected with the Universe. Modern physics views the Universe as a boundless, indivisible network of dynamic activity. Everything in this world is interconnected and influences each other. The influence of higher-state matter on human intellect and earthly processes should be considered as causal phenomena of the Universe.
Abstract: Intelligence, consciousness, knowledge and skills are key concepts related to understanding human activity and cognitive processes. Consciousness is a subjective perception and awareness of the surrounding world, one's sensations, thoughts and feelings. It allows a person to feel like a separate person, to be aware of their actions and experiences....
Show More
-
Research Article
A Fuzzy Neural Network System for Denoising Magnetic Resonance Images
Shubhajoy Das*,
Debashis Das
Issue:
Volume 11, Issue 2, December 2025
Pages:
58-65
Received:
20 September 2025
Accepted:
5 October 2025
Published:
28 October 2025
Abstract: Image acquisition is an essential step in image processing. When the image acquisition is done the image that is generated is subjected to Impulse noise, Gaussian noise etc. We have performed the image denoising on images inflicted with impulse noise. Image denoising is an essential step in all types of image processing. Traditional techniques reduce the noise in the image but it also reduces the quality of the image. Traditional filters like gaussian filter, median filter is analyzed which work in the spatial domain and filters working in the frequency domain are also considered like Butterworth filters, Weiner filter. A Deep residual Neural Network filter is proposed which is compared with the Fuzzy Neural Network denoiser. Their performance is compared on the metrics PSNR and SSIM. The Fuzzy Neural Network system improves the SSIM significantly compared to a deep residual neural network and a comparison is made with traditional image denoising methods. We also compare the performance of the deep residual neural network, Fuzzy Neural Network system and Median denoising algorithm on impulse noise has been compared. The performance of deep neural networks depends on the total number of examples used and the performance can be improved if we have more image pairs.
Abstract: Image acquisition is an essential step in image processing. When the image acquisition is done the image that is generated is subjected to Impulse noise, Gaussian noise etc. We have performed the image denoising on images inflicted with impulse noise. Image denoising is an essential step in all types of image processing. Traditional techniques redu...
Show More
-
Research Article
Improving the Efficiency of Allometric Equations Using Artificial Neural Networks in Coppicing Stands of Brant's Oak
Saman Fallah*,
Javad Soosani,
Hamed Naghavi,
Mohsen Yousofvand Mofrad
Issue:
Volume 11, Issue 2, December 2025
Pages:
66-87
Received:
12 August 2025
Accepted:
26 August 2025
Published:
3 December 2025
DOI:
10.11648/j.ajnna.20251102.14
Downloads:
Views:
Abstract: Estimation of forest trees biomass for various purposes is fundamental and substantia l. One method of estimating biomass uses allometric equations that limit the normality of variables and the homogeneity of variances. In this study, artificial neural networks were used as an alternative method to increase biomass estimation accuracy. Fifty three sprout chumps of Brant's Oak (Quercus brantii Lindl) were randomly selected from the Melah Shabanan of Khorramabad in Iran. Diameter at knee height, diameter at breast height, crown diameter, number of sprouts, and height of trees were measured. To calculate the dry weight of the biomass, adisk3-5cm from the trunk and crown was separated and weighed, and with the ratio of dry weight to fresh weight, the dry weight of the crown, trunk, and aboveground biomass of the trees was calculated. Modeling the relationships between variables with linear and nonlinear regression equations and Multilayer Perceptron and Radial Basis Function neural networks showed that both neural networks could increase the coefficient of determination to R2=0.98 and R2=0.96 and reduce the error to RMSE%=11.6 and RMSE%=16.9 and thus the neural network models can increase the quality forest biomass estimates are compared with allometric equations.
Abstract: Estimation of forest trees biomass for various purposes is fundamental and substantia l. One method of estimating biomass uses allometric equations that limit the normality of variables and the homogeneity of variances. In this study, artificial neural networks were used as an alternative method to increase biomass estimation accuracy. Fifty three ...
Show More
