Gradient-based training and pruning of radial basis function networks with an application in materials physics

Määttä, Jussi; Bazaliy, Viacheslav; Kimari, Jyri; Djurabekova, Flyura; Nordlund, K.; Roos, Teemu

doi:10.1016/j.neunet.2020.10.002

Cited by 15 publications

(10 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[38][39][40] It is used in a variety of disciplines, including mathematical research and the estimation of physical and chemical property values. [41][42][43][44][45] In term of local adaptation and mutual coverage, RBF simulates the receptor domain of the neural network in the human brain. It uses exponential decay functions, such as the Gaussian function, to locally mimic the nonlinear input-output mapping.…”

Section: Mlp-annmentioning

confidence: 99%

“…Radial basis function (RBF) networks, introduced in the 1991s, provide a reliable way to accurately simulate complex problems from multiple input–output data sets 38–40 . It is used in a variety of disciplines, including mathematical research and the estimation of physical and chemical property values 41–45 . In term of local adaptation and mutual coverage, RBF simulates the receptor domain of the neural network in the human brain.…”

Section: Machine Learning Modelmentioning

confidence: 99%

See 1 more Smart Citation

Modeling of the biosorption of Fe(III) ions by olive‐stone activated carbon

Massaoudi,

Echouchene,

Ayed

et al. 2024

Applied Organom Chemis

View full text Add to dashboard Cite

This study deals with the optimization of Fe(III) ion removal using activated carbon from olive stone waste using advanced machine learning models. The main objective is to evaluate and compare the performance of machine learning models, specifically multilayer perceptron artificial neural network (MLP‐ANN), general regression artificial neural network (GR‐ANN), radial basis function artificial neural network (RBF‐ANN), and particle swarm optimization artificial neural network (PSO‐ANN) in predicting Fe(III) removal efficiency. Experimental data on adsorption parameters were used to train and test the models. Techniques such as tuning hidden layer neurons, optimizing propagation values, and using a Taguchi approach PSO algorithm were applied to improve the models. For the MLP‐ANN model, the optimal configuration contains 13 neurons in the hidden layer. Concerning the parameters involved in the PSO‐ANN model, the coefficient C2 and the particle have the main effect on the reduction of the error. Their contributions are respectively 49% and 19%. The PSO‐ANN model showed superior performance with the highest regression coefficient (0.9997) and remarkable prediction accuracy, surpassing other models such as MLP‐ANN and GR‐ANN. This research suggests that innovative optimization techniques, particularly using PSO algorithms, significantly enhance the predictive capabilities of machine learning models in complex adsorption processes, contributing to more accurate Fe(III) removal models.

show abstract

Section: Mlp-annmentioning

confidence: 99%

Section: Machine Learning Modelmentioning

confidence: 99%

Modeling of the biosorption of Fe(III) ions by olive‐stone activated carbon

Massaoudi,

Echouchene,

Ayed

et al. 2024

Applied Organom Chemis

View full text Add to dashboard Cite

show abstract

“…It is used in various fields such as mathematical studies and the estimation of properties in physics and chemistry. [29][30][31][32][33] The RBF models the receptor domain of the human brain's neural network with local adjustment and mutual coverage. To locally approximate nonlinear input-output mapping, it makes use of exponential decay functions like the Gaussian function.…”

Section: Radial Basis Function Neural Network Modelingmentioning

confidence: 99%

“…Radial basis function neural network (RBFNN), introduced in the 1991s, is considered one of the most popular ANN algorithms. It is used in various fields such as mathematical studies and the estimation of properties in physics and chemistry 29–33 . The RBF models the receptor domain of the human brain's neural network with local adjustment and mutual coverage.…”

Section: Theorymentioning

confidence: 99%

Multiple linear regression and artificial neural networks for highly selective cationic β‐diimine‐methallyl nickel (II) catalyst for styrene dimerization reaction to 1,3‐diphenyl‐1‐butene

Landolsi

Echouchene

Bajahzar

et al. 2023

Applied Organom Chemis

View full text Add to dashboard Cite

The objective of the present study is to develop predictive models to input and output parameters for linear styrene dimerization reactions. These reactions involve the conversion of two styrene molecules into 1,3‐diphenyl‐1‐butene, which is a commonly used intermediate in the production of various industrial chemicals. Multiple linear regression (MLR) and artificial neural network based on multilayer perceptron (MLP) and radial basis function were used to model 1,3‐diphenyl‐1‐butne dimerization process in order to evaluate its performance. The neural network has been trained and tested by experimental data. The effect of various parameters (such as complex concentration) on styrene conversion (Conv%) and turnover of frequency (TOF) has been investigated in the proposed work. The results found by the proposed predictive models were analyzed and compared with the experimental results. Based on the comparison of the results, the radial basis function neural network (RBFNN) model outperformed the other models (MLR and MLP) with a correlation coefficient of (0.987 for Conv% and 0.947 for TOF) and lower root mean square errors for the output parameters. This result demonstrates that the RBFNN is an efficient technique to predict the styrene conversion and turnover frequency of the dimerization reaction. It was exposed that the control strategies learned are robust and can be transferred to similar dimerization reaction configurations.

show abstract

“…Consequently, the majority of state-of-the-art machine learning algorithms lack robustness in predicting these systems. Upon availability of sufficient data, these have also garnered considerable success in problems governed by physics, such as dynamical systems (Dana and Wheeler, 2020), geosciences (DeVries et al, 2018;Bergen et al, 2019;Racca and Magri, 2021;Saha et al, 2021;Jahanbakht et al, 2022), material science and informatics (Butler et al, 2018;Ramprasad et al, 2017;Batra et al, 2021;Määttä et al, 2021), fluid mechanics (Kutz, 2017;Brunton et al, 2020), various constitutive modeling (Tartakovsky et al, 2018;Xu et al, 2021), etc. Their applicability however may be further enhanced by utilizing physical information available by mathematical/ analytical means.…”

Section: Introductionmentioning

confidence: 99%

Physics-aware deep learning framework for linear elasticity

Roy¹,

Bose²

2023

Preprint

View full text Add to dashboard Cite

The paper presents an efficient and robust data-driven deep learning (DL) computational framework developed for linear continuum elasticity problems. The methodology is based on the fundamentals of the Physics Informed Neural Networks (PINNs). For an accurate representation of the field variables, a multi-objective loss function is proposed. It consists of terms corresponding to the residual of the governing partial differential equations (PDE), constitutive relations derived from the governing physics, various boundary conditions, and data-driven physical knowledge fitting terms across randomly selected collocation points in the problem domain. To this end, multiple densely connected independent artificial neural networks (ANNs), each approximating a field variable, are trained to obtain accurate solutions. Several benchmark problems including the Airy solution to elasticity and the Kirchhoff-Love plate problem are solved. Performance in terms of accuracy and robustness illustrates the superiority of the

show abstract

Gradient-based training and pruning of radial basis function networks with an application in materials physics

Cited by 15 publications

References 19 publications

Modeling of the biosorption of Fe(III) ions by olive‐stone activated carbon

Modeling of the biosorption of Fe(III) ions by olive‐stone activated carbon

Multiple linear regression and artificial neural networks for highly selective cationic β‐diimine‐methallyl nickel (II) catalyst for styrene dimerization reaction to 1,3‐diphenyl‐1‐butene

Physics-aware deep learning framework for linear elasticity

Contact Info

Product

Resources

About