A deep convolutional fuzzy system (DCFS) on a high-dimensional input space is a multi-layer connection of many low-dimensional fuzzy systems, where the input variables to the low-dimensional fuzzy systems are selected through a moving window across the input spaces of the layers. To design the DCFS based on input-output data pairs, we propose a bottom-up layer-by-layer scheme. Specifically, by viewing each of the first-layer fuzzy systems as a weak estimator of the output based only on a very small portion of the input variables, we design these fuzzy systems using the WM Method. After the first-layer fuzzy systems are designed, we pass the data through the first layer to form a new data set and design the second-layer fuzzy systems based on this new data set in the same way as designing the first-layer fuzzy systems. Repeating this process layer-by-layer we design the whole DCFS. We also propose a DCFS with parameter sharing to save memory and computation. We apply the DCFS models to predict a synthetic chaotic plus random time-series and the real Hang Seng Index of the Hong Kong stock market.

Fast Training Algorithms for Deep Convolutional Fuzzy Systems with Application to Stock Index Prediction

A deep convolutional fuzzy system (DCFS) on a high-dimensional input space is a multilayer connection of many low-dimensional fuzzy systems, where the input variables to the low-dimensional fuzzy systems are selected through a moving window across the input spaces of the layers. To design the DCFS based on an input–output data pairs, we propose a bottom–up layer-by-layer scheme. Specifically, by viewing each of the first-layer fuzzy systems as a weak estimator of the output based only on a very small portion of the input variables, we design these fuzzy systems using the Wang–Mendel method. After the first-layer fuzzy systems are designed, we pass the data through the first layer to form a new dataset and design the second-layer fuzzy systems based on this new dataset in the same way as designing the first-layer fuzzy systems. Repeating this process layer-by-layer, we design the whole DCFS. We also propose a DCFS with parameter sharing to save memory and computation. We apply the DCFS models to predict a synthetic chaotic plus random time-series and the real Hang Seng Index of the Hong Kong stock market.

Fast Training Algorithms for Deep Convolutional Fuzzy Systems With Application to Stock Index Prediction

In this paper, Face recognition system using improved artificial bee colony algorithm is proposed. This system works on two different phases, named as Extraction phase and Recognition phase. In the extraction phase, sub-window extraction algorithm is applied on different face images collected from different sources. Then applied artificial bee colony algorithm to reduce number of sub windows, taken from sub window extraction algorithm. This proposed work helps to improve the overhead problem of PCA algorithm. Experiments are carried out on different datasets, collected from FACE_94, and VITM datasets. Experimental results show that proposed work efficiency, in terms of recognition rate, is better than original PCA algorithm by reducing number of subwindows.

Face recognition system using improved artificial bee colony algorithm

Digital image processing is one of the key feature for security, authentication and authorization over the image. Image segmentation is an essential step to evaluate images and extract data from them. The separation of an image into significant structures, image segmentation is frequently an critical step in image analysis, visualization, object representation and many other image processing tasks. To identify and locating pointed discontinuities in an image, edge detection is needed. It is needed for image analysis for solving many complex problems. It also provides important image information can be used for image interpretation. In this work, firstly all the colored face images are converted into gray level images then apply connected component based segmentation algorithm on all gray level images to extract segments which are most suitable for recognition. In this segmentation algorithm, sobel edge detection algorithm and median filter for noise removal is used. Finally neural network is used to recognize the face images by taking suitable features (like standard deviation, mean gray value, center of mass and integrated density). Experiments are performed on IIITM_Gwalior, VITM dataset and FACE_94 dataset and results show that proposed algorithm recognition rate is better than PCA algorithm.

An improved connected component based algorithm for face recognition

The BP neural network is easy to fall into local minimum point, the algorithm convergence speed slow, this paper puts forward an improved algorithm of flexible neural tree, introduced the basic theory knowledge of Flexible neural tree , analyzes the characteristics and advantages of the neural tree. The structure optimization and parameter optimization are adopted some optimization algorithm, Introduced the multi expression programming algorithm for optimization of flexible neural tree structure and by using the improved particle swarm algorithm to optimize the parameters of flexible neural tree, Finally the establishment of complete flexible neural tree model.

/pdf/study-on-improved-flexible-neural-tree-optimization-4nfor2juxh.pdf

Study on Improved Flexible Neural tree Optimization Algorithm

A new approach for face recognition based on Principal Component Analysis(PCA) and neural networks is(proposed).First the PCA is used to obtain the best description features over the entire image.It can reduce the input of the neural networks.Then the reduced image data is delivered to a forward neural network for training.The weights of the network is optimized by using the particle swarm optimization(PSO).This method was tested by using the standard ORL face database.The results show that the method is more effective than the others.

Face recognition based on PCA and neural network

A combination approach of a modified PIPE algorithm to find the best architecture of hierarchical TS fuzzy model for a given problem and a simple evolutionary programming algorithm to optimize the parameters embedded in the hierarchical TS fuzzy model is proposed. Simulation result for the chaotic time-series prediction problem shows the feasibility and effectiveness of the proposed method.

http://ieeexplore.ieee.org/iel5/9834/30994/01441619.pdf

Approach for evolving hierarchical T-S fuzzy systems

A new neural tree for modeling the time-series forecasting is proposed in the paper.Grammar guided genetic programming was used to optimize the structure of neural tree for modeling the time-series forecasting in this paper,and we compared the performance with an artificial(network) model for time-(series) forecasting.The result shows that our model has a higher reliability than the artificial network model.

Chen Yue-hui

Papers

Face recognition based on PCA and neural network

Approach for evolving hierarchical T-S fuzzy systems

Time-Series Forecasting Based on Neural Tree Model