Article
Computer Science, Artificial Intelligence
Nicolas Couellan
Summary: This study focuses on the robustness of deep neural networks, proposing a simple concentration inequality to study the probability of network output deviating from its nominal value and using network conditions to regularize the loss function. Empirical evaluation shows that the proposed method accurately estimates the observed robustness.
Article
Computer Science, Hardware & Architecture
Dev Narayan Yadav, Phrangboklang Lyngton Thangkhiew, Kamalika Datta, Sandip Chakraborty, Rolf Drechsler, Indranil Sengupta
Summary: This paper introduces a training algorithm based on resistive memory systems that achieves accuracy similar to existing algorithms, but with faster training speed, by using additional memristors and a threshold gate.
JOURNAL OF SYSTEMS ARCHITECTURE
(2022)
Article
Mathematical & Computational Biology
Hacer Turgut, Beste Turanli, Betuel Boz
Summary: Circular RNAs play critical roles in biological processes and are related to human diseases. Computational methods can predict the associations between circular RNAs and diseases, overcoming the limitations of traditional experimental methods. This study proposes a deep learning method that successfully predicts the circular RNA-disease associations.
INTERDISCIPLINARY SCIENCES-COMPUTATIONAL LIFE SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Shaokai Zhao, Bin Chen, Hui Wang, Zhiyuan Luo, Tao Zhang
Summary: A novel feed-forward neural network inspired by the structure of the dentate gyrus and neural oscillatory analysis has been proposed to increase the storage capacity of Hopfield network. By using a mouse model of environmental enrichment and neural oscillatory analysis, a computational model of the dentate gyrus associated with better pattern separation ability has been obtained. The simulation results show significant expansion in storage capacity and improved pattern separation compared to the standard Hopfield network.
INTERNATIONAL JOURNAL OF NEURAL SYSTEMS
(2022)
Article
Chemistry, Physical
Jeremiasz Pilarz, Ilya Polishuk, Miroslaw Chorazewski
Summary: Ionic liquids have great potential as new hydraulic oils and high-pressure heat transfer media. This study proposes a simple neural network model for predicting the speed of sound and adiabatic compressibility data in ionic liquids, and compares its results with the predictions of the CP-PC-SAFT Equation of State.
JOURNAL OF MOLECULAR LIQUIDS
(2022)
Article
Computer Science, Information Systems
Xiaoxiang Guo, Weimin Han, Jingli Ren
Summary: Analysis and prediction of time series are crucial in scientific fields such as meteorology, epidemiology, and economy. This paper proposes a prediction system based on a dynamic feed-forward neural network, utilizing trajectory information in the reconstructed phase space to establish the prediction model. An integer constrained particle swarm optimization algorithm is employed for selecting the optimal time delay parameter. Simulation results on various datasets validate the efficiency and reliability of the proposed method.
SCIENCE CHINA-INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
S. M. Mallikarjunaiah
Summary: This paper explores the use of a deep learning feedforward artificial neural network as a numerical tool for approximating the solutions to singularly perturbed delay differential equations. The approach trains the network with fewer uniform data points using linear interpolation and does not rely on the exact solution for training. The results demonstrate that the fine-tuned adaptive deep learning architecture can effectively approximate the solution to SPDDE for various delay and perturbation parameters.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Rajesh Chatterjee, Ranapratap Mukherjee, Provas Kumar Roy, Dinesh Kumar Pradhan
Summary: The study aims to find an optimization technique to improve the performance of the feed forward neural network (FNN). A new optimization approach called chaotic oppositional-based whale optimization algorithm (COWOA) is developed and compared with other methods for validation.
Article
Chemistry, Multidisciplinary
Tuan-Ho Le, Li Dai, Hyeonae Jang, Sangmun Shin
Summary: This study proposes a neural network-based robust design modeling method for estimating functional relationships between input factors and output responses. Comparative analysis shows that the proposed method has significant advantages in reducing quality loss and variability indicators.
APPLIED SCIENCES-BASEL
(2022)
Article
Environmental Sciences
Nereida Rodriguez-Alvarez, Joseph S. Jao, Joan Francesc Munoz-Martin, Clement G. Lee, Kamal Oudrhiri
Summary: This manuscript investigates the application of a feed-forward neural network denoising methodology to Near-Earth Asteroid (NEA) data obtained from the Goldstone Solar System Radar (GSSR). The results show an increase in signal level and improved radar detection of NEAs when applying the denoising technique. Reducing noise level also benefits shorter integration times and allows for a higher number of independent pieces of information to be captured during antenna beam crossing.
Article
Thermodynamics
Pengyue Wang, Maozu Guo, Yingeng Cao, Shimeng Hao, Xiaoping Zhou, Lingling Zhao
Summary: This study proposes a novel approach, SFGAN, for predicting pedestrian wind flow using generative adversarial networks. By embedding spatial and frequency characteristics, the method enhances predictions and reduces errors compared to previous methods.
BUILDING SIMULATION
(2023)
Article
Computer Science, Interdisciplinary Applications
Dulguun Narmandakh, Christoph Butscher, Faramarz Doulati Ardejani, Huichen Yang, Thomas Nagel, Reza Taherdangkoo
Summary: This article presents the use of neural network models to predict the swelling potential of clay soils, including both natural and artificial soils. The models were trained using the Levenberg-Marquardt algorithm and validated with experimental data, showing that the feed-forward neural network trained with this algorithm is the most accurate.
COMPUTERS AND GEOTECHNICS
(2023)
Article
Materials Science, Multidisciplinary
Guoqing Jing, Lizhen Chen, Peipei Wang, Wenjie Xiong, Zebin Huang, Junmin Liu, Yu Chen, Ying Li, Dianyuan Fan, Shuqing Chen
Summary: A method for recognizing FOAM modes using a feedforward neural network (FNN) is proposed in this study, enabling accurate identification of FOAM modes in turbulent environments. By employing diffraction preprocessing, the FNN is provided with more feature information, extending the detection range for conjugate FOAM modes.
RESULTS IN PHYSICS
(2021)
Article
Engineering, Civil
Yue Pan, Xiankui Zeng, Hongxia Xu, Yuanyuan Sun, Dong Wang, Jichun Wu
Summary: Systematic model error in groundwater predictions can be corrected using Gaussian process regression (GPR) to build a statistical complementary model. The choice of kernel function in GPR has a significant impact on capturing systematic prediction errors, with RQ kernel showing the best performance in improving groundwater predictions among the nine kernels studied.
JOURNAL OF HYDROLOGY
(2021)
Article
Green & Sustainable Science & Technology
Ozge Cagcag Yolcu, Fulya Aydin Temel, Ayse Kuleyin
Summary: In this study, hybrid prediction models were used to estimate the adsorption of ammonium from landfill leachate by using zeolite in batch and column systems. The models showed significant improvement compared to traditional methods, with prediction errors being very low. The findings suggest that the proposed model can be effectively and reliably used without the need for additional experiments in environmental sciences.
JOURNAL OF CLEANER PRODUCTION
(2021)