China
Africa
Explore chapters and articles related to this topic
Prediction of resilient modulus of cohesive subgrade soils from CPTU data using polynomial neural networks
Published in Guido Gottardi, Laura Tonni, Cone Penetration Testing 2022, 2022
Wei Duan, Zening Zhao, Guojun Cai*, Anhui Wang, Ruifeng Chen, Anand J. Puppala, Songyu Liu, Surya Sarat Chandra Congress
With recent advances in computational software and hardware (Zhao et al., 2021a), the use of the artificial neural network (ANN) method as a statistical regression technique to approximate input-output relationships in geotechnical engineering has gained impetus for its ability to effectively deal with the complex relationships in recent years (Juang et al., 1999; Zhao et al., 2021b). However, the major drawback of ANN is that the detected dependencies are hidden in the ANN structure, and the correlations are often not expressed intuitively (Duan et al., 2021a). The group method of data handling (GMDH) type neutral network is a polynomial neural network that using a powerful identification technique to model complex relationships between multiple variables (Moayed et al., 2017). Hence, in the present study, the GMDH model is proposed for predicting Mr based on these parameters of cone tip resistance (qt), sleeve frictional resistance (fs), dry density (γd) and moisture (w).
Ontology Application to Constructing the GMDH-Based Inductive Modeling Tools
Published in Archana Patel, Narayan C. Debnath, Bharat Bhushan, Semantic Web Technologies, 2023
Halyna Pidnebesna, Volodymyr Stepashko
Ivakhnenko has published his first article on the GMDH in 1968 in Ukrainian journal “Avtomatyka” reprinted in the USA [18]. Articles [24–26] published abroad were probably the most known sources on this method. A typical inductive modeling task reduces to automated building a model describing unknown functioning patterns of an object or process on the basis of a given data set. GMDH is an original method for inductive self-organizing construction of models from experimental data under conditions of uncertainty. To solve two optimization tasks, model structure specification (discrete task) and model parameters estimation (continues task), the following principles are applied: generation of models of diverse complexity;inductive process of constructing models with the gradual complication of structures;quality assessment of a model using so-called external criteria based on a dataset division into two subsets: the first for estimating parameters and the second to calculate criteria;non-definitive decisions when evolving models: a subset of the best model is selected at any step of the process;successive selection of best models according to the criteria minima.
Performance evaluation of tunnel type sediment excluder efficiency by machine learning
Published in ISH Journal of Hydraulic Engineering, 2022
N K. Tiwari, Parveen Sihag, Dibyendu Das
Recently, several artificial intelligence methods such as artificial neural network (ANN), fuzzy logic (FL), hybrid adaptive neuro-fuzzy inference system (ANFIS); support vector method (SVM), PSO and machine learning technique were applied to generate the modelling of problems in sediment removal efficiency prediction (Singh et al. 2015, 2016; Singh 2016; Tiwari et al. 2018, 2019). Among these techniques, the GMDH network is known as a system identification method, which is used in many fields in order to model and estimate the behaviours of unknown or complex systems based on given input–output data pairs (Amanifard et al. 2008). GMDH is a learning machine based on the principle of heuristic self-organizing. Past researches showed that GMDH networks could be used as a rich tool and technique in the modelling of environmental processes (Najafzadeh and Azamathulla 2013).
Intelligent step-length adjustment for adaptive path-following in nonlinear structural mechanics based on group method of data handling neural network
Published in Mechanics of Advanced Materials and Structures, 2022
Ali Maghami, Seyed Mahmoud Hosseini
GMDH is a kind of inductive algorithm for computer-based mathematical modeling of multi-parametric datasets that has been applied in a great variety of areas for deep learning, knowledge discovery, data mining, and forecasting [45]. GMDH-based neural network has the ability of self organization, and also works very well as a forecasting method of nonlinear systems [46]. For a nonlinear system with as the output variable, we consider the following relation: where, are input variables. The general connection between inputs and output variables can be determined by a gradually complicated Kolmogorov-Gabor polynomial [44]: in which, M is the number of input variables, and are the coefficients. As the number of inputs or the degree of the polynomial increases, the total number of parameters in the polynomial representation grows rapidly. However, each neuron of GMDH-based network has only two input variables and the quadratic form are considered as follows [47]:
Modeling temperature dependency of oil - water relative permeability in thermal enhanced oil recovery processes using group method of data handling and gene expression programming
Published in Engineering Applications of Computational Fluid Mechanics, 2019
Nait Amar Menad, Zeraibi Noureddine, Abdolhossein Hemmati-Sarapardeh, Shahaboddin Shamshirband, Amir Mosavi, Kwok-wing Chau
Group Method of Data Handling (GMDH) known also as polynomial neural network is one of the most promising families of artificial neural networks (ANNs) (Dargahi-Zarandi, Hemmati-Sarapardeh, Hajirezaie, Dabir, & Atashrouz, 2017). Beside the reliability shown by GMDH in modeling complex systems, it ensures the advantage of providing user-friendly polynomial formula to the system being studied. The conception of GMDH technique consists in employing multiple nodes which belong to intermediate layers. The generated value by each GMDH node is calculated based on a quadratic polynomial model that includes the previous neuron. This GMDH version corresponds to the earliest model that was introduced by (Ivakhnenko, Krotov, & Ivakhnenko, 1970). As the earliest version of GMDH presented some generalization lacks, a modified version, known also as hybrid version, was proposed as an extensive version that includes more interactions between the nodes and variables; hence, this version ensures more flexibility for modeling more complex systems (Rostami et al., 2019). The GMDH hybrid version follows the below-shown rule: where stand for the inputs and output parameters of the model, respectively; denote the polynomial coefficients; and mean respectively, the size of layers and the input parameters number.