Stochastic gradient based extreme learning machines for stable online learning of advanced combustion engines

Janakiraman, Vijay Manikandan; Nguyen, XuanLong; Assanis, Dennis N.

doi:10.1016/j.neucom.2015.11.024

Cited by 44 publications

(28 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The analysis of the stability is divided into two steps. At the end, having been extracted L 2 gains, the largest gain is obtained as the L 2 gain of system (15)- (16). Therefore, the whole system (15)-(16) would be output strictly passive.…”

Section: Belm Stabilitymentioning

confidence: 99%

“…Therefore, the whole system (15)-(16) would be output strictly passive. In the second step, the L 2 finite-time stability for each (19) and (20) is proved by considering (16) as the output. At the end, having been extracted L 2 gains, the largest gain is obtained as the L 2 gain of system (15)- (16).…”

Section: Belm Stabilitymentioning

confidence: 99%

“…While the method has been applied to a sensorless nonlinear closed-loop system, the speed of the estimation has been remarkably fast. Javakiraman et al 16 have developed an algorithm mixed of stochastic gradient and extreme learning machine for the systems that their parameters need to be estimated online. The Lyapunov approach has been used for analyzing the asymptotic stability of the estimation error and the determination of the upper bound for the learning rate, leading to stability learning.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Online and stable parameter estimation based on normalized brain emotional learning model (NBELM)

Akhormeh

Roshanian

MoradiMaryamnegari

et al. 2019

Adaptive Control & Signal

View full text Add to dashboard Cite

Recently, a computational model of Amygdala based on the brain emotional learning is presented by psychologists. This brain emotional learning model (BELM) that has a neuro-inspired architecture is utilized to train the weights which are in Amygdala and Orbitofrontal. In this paper, unknown parameters of dynamic systems are estimated by developing the normalized BELM (NBELM).To this end, after proving the stability of the model output, the sufficient condition for weights convergence is extracted while the sensitivity analysis is applied for this model. In order to evaluate the performance of NBELM, in the first example, the matrices of a twin rotor MIMO system are estimated and compared with the equation error method (EEM). In the second example, the nonlinear model of a servomotor is utilized as a case study. In the third example, the performance of the NBELM in experimental systems is validated using a reaction wheel with a DC motor. An important feature of the brain emotional system is its fast response, leading the NBELM to have a high speed performance in estimating the parameters of dynamic systems. A few number of adjustable parameters and low computing complexity also cause the NBELM to be an appropriate method for online estimation of the unknown parameters of dynamic systems. KEYWORDSAmygdala computing model, BEL model, dynamic system parameter estimation, online and stable estimation Parameters of a large number of industrial systems need to be monitored online for various reasons, ranging from adaptive controls 1 to fault detection systems. 2,3 As a consequence, many research studies have been devoted to introducing novel parameters identification algorithms able to insure the stability of the system. Additionally, the proposed techniques must converge to real values as fast as possible in order to be applicable to engineering systems while high accuracy in the converging process is an evitable need in any novel method. The low complexity of the new algorithm is another challenge, leading the system identification method to be implemented with less effort and become wider. For mentioned challenges to be addressed, various investigations create new frameworks in terms of problem formulation, methodology development, theoretical foundation, software tools, and hardware modules. 4Int J Adapt Control Signal Process. 2019;33:1047-1065.wileyonlinelibrary.com/journal/acs

show abstract

Section: Belm Stabilitymentioning

confidence: 99%

Section: Belm Stabilitymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Online and stable parameter estimation based on normalized brain emotional learning model (NBELM)

Akhormeh

Roshanian

MoradiMaryamnegari

et al. 2019

Adaptive Control & Signal

View full text Add to dashboard Cite

show abstract

“…[21,22]. Other popular passive technologies rely on online learning schemes, in particular online gradient descent, which has been incorporated into drift learning strategies for the simple perceptron, neural networks, or extreme learning machines, as an example [23,24]. The behavior of such models varies extensively across different learning scenarios [11].…”

Section: Models Of On-line Learning Under Concept Driftmentioning

confidence: 99%

Statistical Mechanics of On-Line Learning Under Concept Drift

Straat¹,

Abadi

Göpfert

et al. 2018

Preprint

View full text Add to dashboard Cite

We introduce a modelling framework for the investigation of on-line machine learning processes in non-stationary environments. We exemplify the approach in terms of two specific model situations: In the first, we consider the learning of a classification scheme from clustered data by means of prototype-based Learning Vector Quantization (LVQ). In the second, we study the training of layered neural networks with sigmoidal activations for the purpose of regression. In both cases, the target, i.e. the classification or regression scheme, is considered to change continuously while the system is trained from a stream of labeled data. We extend and apply methods borrowed from statistical physics which have been used frequently for the exact description of training dynamics in stationary environments. Extensions of the approach allow for the computation of typical learning curves in the presence of concept drift in a variety of model situations. First results are presented and discussed for stochastic drift processes in classification and regression problems. They indicate that LVQ is capable of tracking a classification scheme under drift to a non-trivial extent. Furthermore, we show that concept drift can cause the persistence of sub-optimal plateau states in gradient based training of layered neural networks for regression.

show abstract

“…It has been suggested 7 that relatively simple models can effectively capture highly complex and structured problems when they are trained on very large data sets. For solving large-scale regression problems 1 feed-forward neural network model with simple architecture and training phase is the extreme learning machine (ELM) introduced by Huang et al 8 ELM has randomly generated hidden neurons and has shown its good performance and fast speed in numerous applications surveyed by Huang et al [9][10][11] ELM-based algorithms have several variations like the error minimized ELM (EM-ELM), 12 the online sequential fuzzy ELM, 13 the optimally pruned ELM, 14 the ridge regression ELM, 15 the regularized basic ELM, 16 the bidirectional ELM, 17 the weighted ELM 18 for imbalanced data, the random projection-based ELM, 19 the multitask ELM for visual tracking, 20 the stochastic gradient-based ELM (SG-ELM), 21 and others.…”

Section: Introductionmentioning

confidence: 99%

Big data regression with parallel enhanced and convex incremental extreme learning machines

Kokkinos

Margaritis

2017

Computational Intelligence

View full text Add to dashboard Cite

This work considers scalable incremental extreme learning machine (I-ELM) algorithms, which could be suitable for big data regression. During the training of I-ELMs, the hidden neurons are presented one by one, and the weights are based solely on simple direct summations, which can be most efficiently mapped on parallel environments. Existing incremental versions of ELMs are the I-ELM, enhanced incremental ELM (EI-ELM), and convex incremental ELM (CI-ELM). We study the enhanced and convex incremental ELM (ECI-ELM) algorithm, which is a combination of the last 2 versions. The main findings are that ECI-ELM is fast, accurate, and fully scalable when it operates in a parallel system of distributed memory workstations. Experimental simulations on several benchmark data sets demonstrate that the ECI-ELM is the most accurate among the existing I-ELM, EI-ELM, and CI-ELM algorithms. We also analyze the convergence as a function of the hidden neurons and demonstrate that ECI-ELM has the lowest error rate curve and converges much faster than the other algorithms in all of the data sets. The parallel simulations also reveal that the data parallel training of the ECI-ELM can guarantee simplicity and straightforward mappings and can deliver speedups and scale-ups very close to linear.

show abstract

Stochastic gradient based extreme learning machines for stable online learning of advanced combustion engines

Cited by 44 publications

References 45 publications

Online and stable parameter estimation based on normalized brain emotional learning model (NBELM)

Online and stable parameter estimation based on normalized brain emotional learning model (NBELM)

Statistical Mechanics of On-Line Learning Under Concept Drift

Big data regression with parallel enhanced and convex incremental extreme learning machines

Contact Info

Product

Resources

About