两两关系马尔科夫网的自适应组稀疏化学习

稀疏化学习能显著降低无向图模型的参数学习与结构学习的复杂性, 有效地处理无向图模型的学习问题. 两两关系马尔科夫网在多值变量情况下, 每条边具有多个参数, 本文对此给出边参数向量的组稀疏化学习, 提出自适应组稀疏化, 根据参数向量的模大小自适应调整惩罚程度. 本文不仅对比了不同边势情况下的稀疏化学习性能, 为了加速模型在复杂网络中的训练过程, 还对目标函数进行伪似然近似、平均场自由能近似和Bethe自由能近似. 本文还给出自适应组稀疏化目标函数分别使用谱投影梯度算法和投 影拟牛顿算法时的最优解, 并对比了两种优化算法进行稀疏化学习的性能. 实验表明自适 应组稀疏化具有良好的性能.     

A parameter adaptive control structure for linear multivariable systems

This paper presents an adaptive control structure which can be used to assign all poles and zeros of a continuous-time linear multivariable system represented by an (m times m) strictly proper transfer matrixT(s), providedT(s)has no right-half plane zeros. The controller parameters can be directly estimated from input-output data. The paper also serves to point out the type of a priori information necessary for multivariable adaptive controller design. This information is a natural extension of that required in the scalar case.     

Modified adaptive discrete control system containing neural estimator and neural controller

In this paper a modified discrete adaptive control system with neural estimator and neural controller is presented. The structure of the adaptive controller is based on the model presented by Etxebarria (Etxebarria V. Adaptive control of discrete systems using neural networks. IEE Proc. Control Theory Application, Vol. 141, No. 4, July, 1995) where the stability of the control procedure is proved. The Widrow–Hoff procedure of learning and the DARMA model is used for identifying and adjustment of neural network parameters, applied to adaptive control of discrete systems. In this paper the procedure of Etxebarria is modified. The learning rate of the neural network is improved and accelerated using the PD, PI and PID input controllers for input neurons. The effect of adding a momentum term (the past record of the learning) to the learning rule of the neural network is studied. The results are compared and discussed using the examples of Etxebarria and two other case studies. The procedure is extended to multi-input multi-output systems and cases studied are simulated.     

Observer-based Hamiltonian identification for quantum systems

A symmetry-preserving observer-based parameter identification algorithm for quantum systems is proposed. Starting with a 2-level quantum system (qubit), where the unknown parameters consist of the atom-laser frequency detuning and coupling constant, we prove an exponential convergence result. The analysis is inspired by Lyapunov and adaptive control techniques and is based on averaging theory. The observer is then extended to the multi-level case where all the atom-laser coupling constants are unknown. The extension of the convergence analysis is discussed through some heuristic arguments. The relevance and the robustness with respect to various noises are tested through numerical simulations.     

Learning with intelligent tutors and worked examples: selecting learning activities adaptively leads to better learning outcomes than a fixed curriculum

The main learning activity provided by intelligent tutoring systems is problem solving, although several recent projects investigated the effectiveness of combining problem solving with worked examples. Previous research has shown that learning from examples is an effective learning strategy, especially for novice learners. A worked example provides step-by-step explanations of how a problem is solved. Many studies have compared learning from examples to unsupported problem solving, and suggested presenting worked examples to students in the initial stages of learning, followed by problem solving once students have acquired enough knowledge. This paper presents a study in which we compare a fixed sequence of alternating worked examples and tutored problem solving with a strategy that adapts learning tasks to students’ needs. The adaptive strategy determines the type of the task (a worked example, a faded example or a problem to be solved) based on how much assistance the student received on the previous problem. The results show that students in the adaptive condition learnt significantly more than their peers who were presented with a fixed sequence of worked examples and problem solving. Novices from the adaptive condition learnt faster than novices from the control group, while the advanced students from the adaptive condition learnt more than their peers from the control group.     

Gradient‐based back‐propagation dynamical iterative learning scheme for the neuro‐fuzzy inference system

In this paper, a gradient‐based back propagation dynamical iterative learning algorithm is proposed for structure optimization and parameter tuning of the neuro‐fuzzy system. Premise and consequent parameters of the neuro‐fuzzy model are initialized randomly and then tuned by the proposed iterative algorithm. The learning algorithm is based on the first order partial derivative of the output with respect to the structure parameters. The first order derivative of the model output with respect to the structure parameters determines the sensitivity of the model to structure parameters. The sensitivity values are then used to set the tuning factors and parameters updating step sizes. Therefore, an adaptive dynamical iterative scheme is achieved which adapts the learning procedure to the current state of the performance during the optimization process. Larger tuning step sizes make the convergence speed higher and vice versa. In this regard, this parameter is treated according to the calculated sensitivity of the model to the parameter. The proposed learning algorithm is compared with the least square back propagation method, genetic algorithm and chaotic genetic algorithm in the neuro‐fuzzy model structure optimization. Smaller mean square error and shorter learning time are sought in this paper, and the performance of the proposed learning algorithm is versified regarding these criteria.     

Turbulent Convection Driven by a Constant Temperature Gradient

To better understand the nature of small-scale fluctuations and spectra in turbulent convection, we consider the theoretically interesting case of buoyancy driven thermal convection without boundary layers; that is we use numerical simulations to study the case of turbulent convection in a periodic box driven by a constant temperature gradient. High Rayleigh numbers are achieved using hyperviscous dissipation. The system develops constant heat flux via transport by a few strong ascending/descending jets. This heat flux is highly intermittent even at integral scales. Also, the heat flux depends only weakly on viscosity. We find that the scaling laws for spectra of temperature and velocity fluctuations are consistent with Kolmogorov scaling and inconsistent with Bolgiano-Obukhov scaling. At the level of spectra, the system is approximately locally isotropic. We also find that constant temperature gradient thermal convection exhibits a unique kind of intermittent structure in that scaling exponents of nth order moments of temperature differences saturate when n is large enough. A comparison is made between constant temperature gradient thermal convection and convection of a passive scalar with a constant scalar gradient. It is shown that at small scales the statistical properties of passive scalar convection are quite similar to that of thermal convection.     

Intelligent control of static synchronous series compensator via an adaptive self-tuning PID controller for suppression of torsional oscillations

In this paper, an intelligent controller is proposed to control a static synchronous series compensator (SSSC) in order to mitigate subsynchronous resonance (SSR) oscillations in a power system. This intelligent controller is an adaptive self-tuning PID controller. To train the PID controller, the gradient descent method is employed where the learning rate is adapted in every iteration in order to accelerate the speed of convergence. This control scheme also requires a wavelet neural network (WNN) to identify the controlled system dynamic. To update the parameters of WNN, the gradient descent (GD) along with the adaptive learning rates derived by the Lyapunov method is used. The computer simulations are used to show the ability of the proposed controller. In addition, the performance of the proposed controller is compared with another self-tuning PID controller. The results demonstrate that the proposed controller has a successful performance in minimizing the SSR.     

An improved adaptive control for robust adaptation

An adaptive control law consisting of an integral and proportional adaptation is suggested. It is shown that the addition of the proportional adaptation term to the control law improves the performance of the adaptive system. Although this is done in the context of a first-order system, it is believed that this may also be the case for higher-order systems. The scalar case is only discussed. Simulation results are presented to complement the theoretical developments     

不确定混沌系统的差分进化小波神经网络控制

提出一种基于小生境自适应差分进化小波神经网络（NADE-WNN）的方法对不确定混沌系统进行控制。该方法利用小波神经网络学习未知模型混沌系统的动态特性并实施控制,为提高神经网络的学习精度和收敛速度,采用小生境自适应差分进化算法同时优化小波神经网络的结构和参数,简化网络结构,提高网络的学习精度和全局收敛性。仿真实验结果表明,在有外部干扰和参数摄动的情况下,NADE-WNN仍能对不确定混沌系统进行有效控制,且网络结构、控制精度和收敛速度都优于传统神经网络。     

On the local stability properties of adaptive parameter estimatorswith composite errors and split algorithms

The stability characteristics of a generalized error system structure are examined for adaptive parameter estimation systems. This error system form encompasses the structure of a number of particular applications of adaptive parameter estimation theory which fall outside the framework of more familiar error system models. The consequences for error system stability which derive from the added generality are analyzed. It is demonstrated that, when using averaging theory techniques to analyze the error system, stability conditions must be augmented for the basic error system in order to ensure stability for the generalized error system. The inadequacy of regressor spectral restrictions and persistent spanning conditions to guarantee local error system stability of the generalized structure is rigorously established. Alternative conditions which will yield such local stability are demonstrated. To illustrate the concepts involved, the recursive identification of parameters is examined in a parallel-form realization of a linear system     

Online adaptive fuzzy neural identification and control of a class of MIMO nonlinear systems

This paper presents a robust adaptive fuzzy neural controller (AFNC) suitable for identification and control of a class of uncertain multiple-input-multiple-output (MIMO) nonlinear systems. The proposed controller has the following salient features: 1) self-organizing fuzzy neural structure, i.e., fuzzy control rules can be generated or deleted automatically; 2) online learning ability of uncertain MIMO nonlinear systems; 3) fast learning speed; 4) fast convergence of tracking errors; 5) adaptive control, where structure and parameters of the AFNC can be self-adaptive in the presence of disturbances to maintain high control performance; 6) robust control, where global stability of the system is established using the Lyapunov approach. Simulation studies on an inverted pendulum and a two-link robot manipulator show that the performance of the proposed controller is superior.     

不确定分数阶多涡卷混沌系统自适应重复学习同步控制

研究了不确定分数阶多涡卷混沌系统的自适应重复学习同步控制问题.通过利用滞环函数,设计了一类参数可调的分数阶多涡卷混沌系统.针对这类分数阶多涡卷混沌系统,在考虑非参数化不确定性、周期时变参数化不确定性、常参数化不确定性和外部扰动情况下,提出了一种重复学习同步控制方案.利用自适应神经网络技术补偿了系统中的函数型不确定性,通过自适应重复学习控制技术处理了周期时变参数化不确定性,并利用自适应鲁棒学习项处理了神经网络逼近误差和干扰的影响,实现了主系统和从系统的完全同步.综合利用分数阶频率分布模型和类Lyapunov复合能量函数方法证明了同步误差的学习收敛性.数值仿真验证了所提方法的有效性.     

Large Scale Nonlinear Control System Fine-Tuning Through Learning

Despite the continuous advances in the fields of intelligent control and computing, the design and deployment of efficient large scale nonlinear control systems (LNCSs) requires a tedious fine-tuning of the LNCS parameters before and during the actual system operation. In the majority of LNCSs the fine-tuning process is performed by experienced personnel based on field observations via experimentation with different combinations of controller parameters, without the use of a systematic approach. The existing adaptive/neural/fuzzy control methodologies cannot be used towards the development of a systematic, automated fine-tuning procedure for general LNCS due to the strict assumptions they impose on the controlled system dynamics; on the other hand, adaptive optimization methodologies fail to guarantee an efficient and safe performance during the fine-tuning process, mainly due to the fact that these methodologies involve the use of random perturbations. In this paper, we introduce and analyze, both by means of mathematical arguments and simulation experiments, a new learning/adaptive algorithm that can provide with convergent, an efficient and safe fine-tuning of general LNCS. The proposed algorithm consists of a combination of two different algorithms proposed by Kosmatopoulos (2007 and 2008) and the incremental-extreme learning machine neural networks (I-ELM-NNs). Among the nice properties of the proposed algorithm is that it significantly outperforms the algorithms proposed by Kosmatopoulos as well as other existing adaptive optimization algorithms. Moreover, contrary to the algorithms proposed by Kosmatopoulos , the proposed algorithm can operate efficiently in the case where the exogenous system inputs (e.g., disturbances, commands, demand, etc.) are unbounded signals.      

An efficient compensatory neuro-fuzzy system and its applications

In this study, a compensatory neuro-fuzzy system (CNFS) is proposed. The compensatory fuzzy reasoning method uses adaptive fuzzy operations of a neuro-fuzzy system to make the fuzzy logic system more adaptive and effective. Furthermore, an online learning algorithm that consists of structure learning and parameter learning is proposed to automatically construct the CNFS. The structure learning is based on the fuzzy similarity measure to determine the number of fuzzy rules, and the parameter learning is based on backpropagation algorithm to adjust the parameters. The simulation results have shown that (1) the CNFS model converges quickly and (2) the CNFS model has a lower root mean square (RMS) error than other models.     

A self‐tuning iterative learning controller for time variant systems

We consider the iterative learning control problem from an adaptive control viewpoint. The self‐tuning iterative learning control systems (STILCS) problem is formulated in a general case, where the underlying linear system is time‐variant and its parameters are all unknown and where its initial conditions are not constant and not determinable in various iterations. A procedure for solving this problem will be presented. The Lyapunov technique is employed to ensure the convergence of the presented STILCS. Computer simulation results are included to illustrate the effectiveness of the proposed STILCS. Copyright © 2008 John Wiley and Sons Asia Pte Ltd and Chinese Automatic Control Society     

MODELING AND CONTROL OF INTERNAL COMBUSTION ENGINES USING INTELLIGENT TECHNIQUES

This article will compare two different fuzzy-derived techniques for controlling small internal combustion engine and modeling fuel spray penetration in the cylinder of a diesel internal combustion engine. The first case study is implemented using conventional fuzzy-based paradigm, where human expertise and operator knowledge were used to select the parameters for the system. The second case study used an adaptive neuro-fuzzy inference system (ANFIS), where automatic adjustment of the system parameters is affected by a neural networks based on prior knowledge. The ANFIS model was shown to achieve an improved accuracy compared to a pure fuzzy model, based on conveniently selected parameters. Future work is concentrating on the establishment of an improved neuro-fuzzy paradigm for adaptive, fast and accurate control of small internal combustion engines.     

Parameter adaptive identification and control

A new parameter adaptive control scheme is outlined which essentially allows one to arbitrarily position all of the poles of an unknown, linear, scalar system of known dynamical order. The technique is based on identifying the parameters which characterize the dynamical behavior of the system in differential operator form and simultaneously implementing an adaptive observer which generates a feedback control signal which converges to the equivalent of an appropriate pole placing linear state variable control law. It is shown that the overall control configuration, consisting of a system identifier and a complementary adaptive observer, is globally stable provided only that the control input is "sufficiently rich" in frequency content. A study employing an unstable, nonminimum phase plant is presented to illustrate the technique.     

A Pareto biogeography-based optimisation for multi-objective two-sided assembly line sequencing problems with a learning effect

This research presents a Pareto biogeography-based optimisation (BBO) approach to mixed-model sequencing problems on a two-sided assembly line where a learning effect is also taken into consideration. Three objectives which typically conflict with each other are optimised simultaneously comprising minimising the variance of production rate, minimising the total utility work and minimising the total sequence-dependent setup time. In order to enhance the exploration and exploitation capabilities of the algorithm, an adaptive mechanism is embedded into the structure of the original BBO, called the adaptive BBO algorithm (A-BBO). A-BBO monitors a progressive convergence metric in every certain generation and then based on this data it will decide whether to adjust its adaptive parameters to be used in the next certain generations or not. The results demonstrate that A-BBO outperforms all comparative algorithms in terms of solution quality with indifferent solution diversification.     

An adaptive ordered fuzzy time series with application to FOREX

An adaptive ordered fuzzy time series is proposed that employs an adaptive order selection algorithm for composing the rule structure and partitions the universe of discourse into unequal intervals based on a fast self-organising strategy. The automatic order selection of FTS as well as the adaptive partitioning of each interval in the universe of discourse is shown to greatly affect forecasting accuracy. This strategy is then applied to prediction of FOREX market. Financial markets, such as FOREX, are generally attractive applications of FTS due to their poorly understood model as well as their great deal of uncertainty in terms of quote fluctuations and the behaviours of the humans in the loop. Specifically, since the FOREX market can exhibit different behaviours at different times, the adaptive order selection is executed online to find the best order of the FTS for current prediction. The order selection module uses voting, statistical analytic and emotional decision making agents. Comparison of the proposed method with earlier studies demonstrates improved prediction accuracy at similar computation cost.