Untitled

Schädler, Kristina; Wysotzki, Fritz

doi:10.1023/a:1008320413168

Cited by 26 publications

(2 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…RNNs can be modified to generate adaptive neural network structures like long–short-term memory (LSTM), which enhances the performance of the RNNs by storing information for long or short durations of time and preventing the problems of vanishing gradients during parameter estimation . While existing adaptive NN models such as adaptive bidirectional associative memory (ABAM) and transversal/recursive filters , have found applications in signal processing and communication, they may be inadequate for many chemical engineering applications where learning must be accomplished with data that may be noisy, limited, and time-varying.…”

Section: Introductionmentioning

confidence: 99%

Hybrid Series/Parallel All-Nonlinear Dynamic-Static Neural Networks: Development, Training, and Application to Chemical Processes

Mukherjee

Bhattacharyya

2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

This paper presents the development of data-driven hybrid nonlinear static-nonlinear dynamic neural network models and addresses the challenges of optimal estimation of parameters for such hybrid networks. A parallel static-dynamic neural network and two variants of series networks, specifically, nonlinear static-nonlinear dynamic and nonlinear dynamic-nonlinear static networks, are investigated in this work. Performances of the proposed fully nonlinear hybrid series and parallel network models are compared with the existing state-of-the-art data-driven models like long−short-term memory networks and gated recurrent unit models as well as DABNet-type linear-dynamic-nonlinear-static neural network. Algorithms are developed for training the series and parallel hybrid networks where the static and dynamic networks can be trained independently by different algorithms. These algorithms offer flexibility for incorporating different types of static and dynamic network architectures and their training algorithms thus offering tradeoff between computational expense and accuracy for highly nonlinear systems. In addition to the typical training objective that minimizes some squared error, an objective function is considered that penalizes overfitting and uncertainty in parameter estimates. Performances of the proposed algorithms are evaluated for three nonlinear example problems�a two-tank pH neutralization reactor, the widely used Van de Vusse reactor, and a pilot plant for postcombustion CO 2 capture using monoethanolamine solvent. Computational expense and convergence performance of the proposed network architectures and their training algorithms are presented.

show abstract

Section: Introductionmentioning

confidence: 99%

Hybrid Series/Parallel All-Nonlinear Dynamic-Static Neural Networks: Development, Training, and Application to Chemical Processes

Mukherjee

Bhattacharyya

2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

show abstract

“…Related approaches have been proposed in the literature (see Kireev and Schädler and Wysotzki); however, no successful methods for training these networks have been proposed. In this work, we propose to employ the well-known back-propagation procedure in combination with stochastic gradient descent to train the proposed network structure in order to perform supervised learning on data sets comprised of molecules and corresponding output properties.…”

Section: Introductionmentioning

confidence: 99%

Automatic Generation of Complementary Descriptors with Molecular Graph Networks

Merkwirth

Lengauer

2005

J. Chem. Inf. Model.

View full text Add to dashboard Cite

We describe a method for the automatic generation of weakly correlated descriptors for molecular data sets. The method can be regarded as a statistical learning procedure that turns the molecular graph, representing the 2D formula of the compound, into an adaptive whole molecule composite descriptor. By translating the molecular graph structure into a dynamical system, the algorithm can compute an output value that is highly sensitive to the molecular topology. This system can be trained by gradient descent techniques, which rely on the efficient calculation of the gradient by back-propagation. We present computational experiments concerning the classification of the Developmental Therapeutics Program AIDS antiviral screen data set on which the performance of the method compares with that of approaches based on substructure comparison.

show abstract

All‐nonlinear static‐dynamic neural networks versus Bayesian machine learning for data‐driven modelling of chemical processes

Mukherjee,

Adeyemo,

Bhattacharyya

2024

Can J Chem Eng

View full text Add to dashboard Cite

In recent decades, the utilization of machine learning (ML) and artificial intelligence (AI) approaches have been explored for process modelling applications. However, different types of ML models may have contrasting advantages and disadvantages, which become critical during the optimal selection of a specific data‐driven model for a particular application as well as estimation of parameters during model training. This paper compares and contrasts two different types of data‐driven modelling approaches, namely the series/parallel all‐nonlinear static‐dynamic neural network models and models from a Bayesian ML approach. Both types of AI modelling approaches considered in this work have shown to significantly outperform several state‐of‐the‐art steady‐state and dynamic data‐driven modelling techniques for various performance measures, specifically, model sparsity, predictive capabilities, and computational expense. The performances of the proposed model structures and algorithms have been evaluated for two nonlinear dynamic chemical engineering systems—a plug‐flow reactor for vapour phase cracking of acetone for production of acetic anhydride and a pilot‐plant for post‐combustion CO2 capture using monoethanolamine as the solvent. For the validation data from the CO2 capture pilot plant, root mean squared error (RMSE) for flue gas outlet temperature, flowrate and CO2 concentration is 0.05%, 1.07%, and 5.0%, respectively, for the all‐nonlinear static‐dynamic neural networks and 0.1%, 1.75%, and 14.14%, respectively, for the Bayesian ML models. For the plug flow reactor data, the Bayesian ML models yield superior RMSE compared to the all‐nonlinear static‐dynamic neural networks when the measurement data are corrupted with Gaussian, auto‐correlated, or cross‐correlated noise.

show abstract

Untitled

Cited by 26 publications

References 55 publications

Hybrid Series/Parallel All-Nonlinear Dynamic-Static Neural Networks: Development, Training, and Application to Chemical Processes

Hybrid Series/Parallel All-Nonlinear Dynamic-Static Neural Networks: Development, Training, and Application to Chemical Processes

Automatic Generation of Complementary Descriptors with Molecular Graph Networks

All‐nonlinear static‐dynamic neural networks versus Bayesian machine learning for data‐driven modelling of chemical processes

Contact Info

Product

Resources

About