-
Robust Exponential Stability of Periodic Solutions for Static Recurrent Neural Networks with Delays
Guang-Hua Zhang,
Hong Zhang,
Jiangfeng Li,
Shanzai Lee
Issue:
Volume 2, Issue 4, December 2017
Pages:
113-118
Received:
6 February 2017
Accepted:
22 May 2017
Published:
14 July 2017
DOI:
10.11648/j.mlr.20170204.11
Downloads:
Views:
Abstract: In this paper, we study the existence of periodic solutions of time-invariant static recurrent neural networks by using the fixed point theory, Poineare map and Lyapunov function combined with inequality techniques. The static recurrent neural network is a kind of neural network which studies the external states of neurons as variables. And its global robust exponential stability. This paper introduces the research status of artificial neural network, summarizes the research background and development of static recurrent neural network dynamic system, and introduces the main work of this paper. Using the fixed point theory, M. The existence of periodic solutions and the global robust exponential stability of the static recursive neural network with variable delays and the existence of almost periodic solutions of the static recursive neural network of the partitioned time are studied by combining the properties of the matrix and the Lyapunov function combined with the inequality technique. Global exponential stability, the stability conditions of the corresponding problem are obtained respectively, and the results of the related research are generalized. Using Lyapunov. The stability of the quasi - static neural recursive neural network and the stability of the periodic solution are studied. The condition of the stationary static recursive neural network is obtained and the correctness of the condition is illustrated. Considering the influence of stochastic perturbation on the dynamic behavior of static recurrent neural network, the static recursive neural network with time delay and the static recursive neural network with distributed time delay are studied by using the infinitesimal operator, Ito formula and the convergence theorem of martingales. Global critical exponential stability of quasi - static neural network with stochastic perturbation. The static recursive neural network with Markovian modulation and the time-delay static recurrent neural network model considering both random perturbation and Markovian switching are studied. The linear matrix inequality, the finite state space Markov chain property and the Lyapunov-krasovskii function, The judgment condition of the global exponential stability of the system is obtained. Firstly, the global exponential stability problem of quasi - static neural neural network with time - delay and recursive neural network is studied by using the generalized Halanay inequality. Then the stability of the Markovian response sporadic static recurrent neural network is studied by combining the properties of Markov chain.
Abstract: In this paper, we study the existence of periodic solutions of time-invariant static recurrent neural networks by using the fixed point theory, Poineare map and Lyapunov function combined with inequality techniques. The static recurrent neural network is a kind of neural network which studies the external states of neurons as variables. And its glo...
Show More
-
Global Exponential Stability of Periodic Solutions for Static Recurrent Neural Networks with Impulsive Finite
Xuan Guo,
Hong Zhang,
Shanzai Lee
Issue:
Volume 2, Issue 4, December 2017
Pages:
119-124
Received:
7 February 2017
Accepted:
22 May 2017
Published:
17 July 2017
DOI:
10.11648/j.mlr.20170204.12
Downloads:
Views:
Abstract: In this paper, we consider the sufficient conditions for the stability of periodic solutions of static recurrent neural networks with impulsive delay. In this paper, we study the time - delay static recurrent neural network affected by pulse. The results show that the neural network is stable when the pulse function is linear and relatively small, and a condition for the periodic solution with exponential stability is obtained. This paper introduces the research status of artificial neural network, summarizes the research background and development of static recurrent neural network dynamic system, and introduces the main work of this paper. Using the fixed point theory, M. The existence of periodic solutions and the global robust exponential stability of the static recursive neural network with variable delays and the existence of almost periodic solutions of the static recursive neural network of the partitioned time are studied by combining the properties of the matrix and the Lyapunov function combined with the inequality technique. Global exponential stability, the stability conditions of the corresponding problem are obtained respectively, and the results of the related research are generalized. Using Lyapunov. The stability of the quasi - static neural recursive neural network and the stability of the periodic solution are studied. The condition of the stationary static recursive neural network is obtained and the correctness of the condition is illustrated. Considering the influence of stochastic perturbation on the dynamic behavior of static recurrent neural network, the static recursive neural network with time delay and the static recursive neural network with distributed time delay are studied by using the infinitesimal operator, Ito formula and the convergence theorem of martingales. Global critical exponential stability of quasi - static neural network with stochastic perturbation. The static recursive neural network with Markovian modulation and the time-delay static recurrent neural network model considering both random perturbation and Markovian switching are studied. The linear matrix inequality, the finite state space Markov chain property and the Lyapunov-krasovskii function, The judgment condition of the global exponential stability of the system is obtained. Firstly, the global exponential stability problem of quasi - static neural neural network with time - delay and recursive neural network is studied by using the generalized Halanay inequality. Then the stability of the Markovian response sporadic static recurrent neural network is studied by combining the properties of Markov chain.
Abstract: In this paper, we consider the sufficient conditions for the stability of periodic solutions of static recurrent neural networks with impulsive delay. In this paper, we study the time - delay static recurrent neural network affected by pulse. The results show that the neural network is stable when the pulse function is linear and relatively small, ...
Show More
-
Unsupervised Dimensionality Reduction for High-Dimensional Data Classification
Issue:
Volume 2, Issue 4, December 2017
Pages:
125-132
Received:
20 July 2017
Accepted:
9 August 2017
Published:
31 August 2017
DOI:
10.11648/j.mlr.20170204.13
Downloads:
Views:
Abstract: This paper carries on research surrounding the influences produced by dimensionality reduction on machine learning classification effect. Firstly, paper constructs the analysis architecture of data dimension reduction classification, combines the two different unsupervised dimension reduction methods, locally linear embedding (LLE) and principal component analysis (PCA) with the five machine learning classification methods: Gradient Boosting Decision Tree (GBDT), Random Forest, Support Vector Machine (SVM), K-Nearest Neighbor (KNN) and Logistic Regression. And then uses the handwritten digital identification dataset to analyze the classification performance of these five classification methods on different dimension datasets by different dimensionality reduction methods. The analysis shows that using the appropriate dimensionality reduction method for dimensionality reduction classification can effectively improve the classification accuracy; the dimensionality reduction classification effect of non-linear dimensionality reduction method is generally better than the linear dimensionality reduction method; different machine learning classification algorithms have significant differences in the sensitivity of dimensions.
Abstract: This paper carries on research surrounding the influences produced by dimensionality reduction on machine learning classification effect. Firstly, paper constructs the analysis architecture of data dimension reduction classification, combines the two different unsupervised dimension reduction methods, locally linear embedding (LLE) and principal co...
Show More
-
Performance Evaluation of Cooperative RL Algorithms for Dynamic Decision Making in Retail Shop Application
Deepak Annasaheb Vidhate,
Parag Arun Kulkarni
Issue:
Volume 2, Issue 4, December 2017
Pages:
133-147
Received:
27 September 2017
Accepted:
20 October 2017
Published:
12 December 2017
DOI:
10.11648/j.mlr.20170204.14
Downloads:
Views:
Abstract: A novel approach by Expertise based Multi-agent Cooperative Reinforcement Learning Algorithms (EMCRLA) for dynamic decision-making in the retail application is proposed in this paper. Performance evaluation between Cooperative Reinforcement Learning Algorithms and Expertise based Multi-agent Cooperative Reinforcement Learning Algorithms (EMCRLA) is demonstrated. Different cooperation schemes for multi-agent cooperative reinforcement learning i.e. EQ learning, EGroup scheme, EDynamic scheme and EGoal driven scheme are proposed here. Implementation outcome includes a demonstration of recommended cooperation schemes that are competent enough to speed up the collection of agents that achieve excellent action policies. This approach is developed for three retailer stores in the retail marketplace. Retailers are able to help with each other and can obtain profit from cooperation knowledge through learning their own strategies that exactly stand for their aims and benefit. The vendors are the knowledgeable agents in the hypothesis to employ cooperative learning to train helpfully in the circumstances. Assuming significant hypothesis on the vendor’s stock policy, restock period, arrival process of the consumers, the approach is modeled as Markov decision process model that makes it possible to design learning algorithms. Dynamic consumer performance is noticeably learned using the proposed algorithms. The paper illustrates results of Cooperative Reinforcement Learning Algorithms of three shop agents for the period of one-year sale duration and then demonstrated the results using proposed approach for three shop agents for the period of one-year sale duration. The results obtained by the proposed expertise based cooperation approach show that such methods can put into a quick convergence of agents in the dynamic environment.
Abstract: A novel approach by Expertise based Multi-agent Cooperative Reinforcement Learning Algorithms (EMCRLA) for dynamic decision-making in the retail application is proposed in this paper. Performance evaluation between Cooperative Reinforcement Learning Algorithms and Expertise based Multi-agent Cooperative Reinforcement Learning Algorithms (EMCRLA) is...
Show More
-
Early Assessment of Mild Alzheimer’s Disease Using Elman Neural Network, LDA and SVM Methods
Peyman Goli,
Elias Mazrooei Rad,
Kavian Ghandehari,
Mehdi Azarnoosh
Issue:
Volume 2, Issue 4, December 2017
Pages:
148-151
Received:
23 October 2017
Accepted:
10 November 2017
Published:
15 December 2017
DOI:
10.11648/j.mlr.20170204.15
Downloads:
Views:
Abstract: This research provides a process for diagnosising the mild Alzheimer's disease from the brain signals. Due to the material and spiritual costs of nursing, carring and treatment of this disease, the early acurate diagnosis would be much usedful. Considering the effect of the mild Alzheimer's disease on electroencephalography (EEG), the mild Alzheimer would be diagnosed within the early steps by an appropriate process. First, the brain signals of healthy people and patients are registered for four states: closed–eyes, opened–eyes, recall and stimulation, in three channels Pz, Cz and Fz. Then, optimal features are drawn out by using an Elman neural network and two claaaifiers applying genetic algorithm: linear discriminant analysis (LDA) and Support vector machine (SVM). According to the results of testing phase, among the three channels and four states, Elman neural network is much more efficient for Alziemer diagnosising in Pz channel and the state of irritation in comparison with LDA and SVM in the other channels and states.
Abstract: This research provides a process for diagnosising the mild Alzheimer's disease from the brain signals. Due to the material and spiritual costs of nursing, carring and treatment of this disease, the early acurate diagnosis would be much usedful. Considering the effect of the mild Alzheimer's disease on electroencephalography (EEG), the mild Alzheime...
Show More
-
Radial Basis Function Neuroscaling Algorithms for Efficient Facial Image Recognition
Vincent A. Akpan,
Joshua B. Agbogun,
Michael T. Babalola,
Bamidele A. Oluwade
Issue:
Volume 2, Issue 4, December 2017
Pages:
152-168
Received:
30 September 2017
Accepted:
9 November 2017
Published:
28 December 2017
DOI:
10.11648/j.mlr.20170204.16
Downloads:
Views:
Abstract: A Radial basis function neural network-based probabilistic principal component analysis (RBFNN-PPCA) on image recognition based on facial recognition was made. The variational properties of face images are investigated with Eigenfaces algorithm to validate the proposed RBFNN-PPCA algorithm and technique for enhanced optimal image recognition system design. Ten different face image samples for each one hundred different individuals with their corresponding bio-data were taken under different light intensities were cropped and pre-processed. The resulting one thousand face image samples were split into 80% as the training set which constitutes the database of known face images and 20% as the test set which constitutes unknown faces images. Analysis was made on the one thousand face images based on the proposed RBFNN-PPCA algorithm and the Eigenfaces algorithm. The two algorithms were applied simultaneously for enhanced optimal face recognition, and the simulation results show that the proposed face image evaluation techniques as well as the proposed RBF neuroscaling algorithm recognizes a known face image or rejects an unknown face based on the database contents to a high degree of accuracy. The proposed face recognition strategy can be adapted for the design of on-line real-time embedded face recognition systems for public, private, business, commercial or industrial applications.
Abstract: A Radial basis function neural network-based probabilistic principal component analysis (RBFNN-PPCA) on image recognition based on facial recognition was made. The variational properties of face images are investigated with Eigenfaces algorithm to validate the proposed RBFNN-PPCA algorithm and technique for enhanced optimal image recognition system...
Show More
