Unsupervised local deep feature for image recognition

An improved fuzzy C-means clustering algorithm using Euclidean distance function

The fuzzy c-mean (FCM) clustering algorithm is a typical algorithm using Euclidean distance for data clustering and it is also one of the most popular fuzzy clustering algorithms. However, FCM does not perform well in noisy environments due to its possible constraints. To improve the clustering accuracy of item varieties, an improved fuzzy c-mean (IFCM) clustering algorithm is proposed in this paper. IFCM uses the Euclidean distance function as a new distance measure which can give small weights to noisy data and large weights to compact data. FCM, possibilistic C-means (PCM) clustering, possibilistic fuzzy C-means (PFCM) clustering and IFCM are run to compare their clustering effects on several data samples. The clustering accuracies of IFCM in five datasets IRIS, IRIS3D, IRIS2D, Wine, Meat and Apple achieve 92.7%, 92.0%, 90.7%, 81.5%, 94.2% and 88.0% respectively, which are the highest among the four algorithms. The final simulation results show that IFCM has better robustness, higher clustering accuracy and better clustering centers, and it can successfully cluster item varieties.

Object-Specific Four-Path Network for Stroke Risk Stratification of Carotid Arteries in Ultrasound Images

Atherosclerotic carotid plaques have been shown to be closely associated with the risk of stroke. Since patients with symptomatic carotid plaques have a greater risk for stroke, stroke risk stratification based on the classification of carotid plaques into symptomatic or asymptomatic types is crucial in diagnosis, treatment planning, and medical treatment monitoring. A deep learning technique would be a good choice for implementing classification. Usually, to acquire a high-accuracy classification, a specific network architecture needs to be designed for a given classification task. In this study, we propose an object-specific four-path network (OSFP-Net) for stroke risk assessment by integrating ultrasound carotid plaques in both transverse and longitudinal sections of the bilateral carotid arteries. Each path of the OSFP-Net comprises of a feature extraction subnetwork (FE) and a feature downsampling subnetwork (FD). The FEs in the four paths use the same network structure to automatically extract features from ultrasound images of carotid plaques. The FDs use different object-specific pooling strategies for feature downsampling based on the observation that the sizes and shapes in the feature maps obtained from FEs should be different. The object-specific pooling strategies enable the network to accept arbitrarily sized carotid plaques as input and to capture a more informative context for improving the classification accuracy. Extensive experimental studies on a clinical dataset consisting of 333 subjects with 1332 carotid plaques show the superiority of our OSFP-Net against several state-of-the-art deep learning-based methods. The experimental results demonstrate better clinical agreement between the ground truth and the prediction, which indicates its great potential for use as a risk stratification and as a monitoring tool in the management of patients at risk for stroke.

Evaluation of scientific and technological achievements in colleges based on machine learning and image feature retrieval

At present, experts and scholars have conducted more research on the ability of colleges and universities to transform scientific and technological achievements. However, they pay more attention to the holistic research on the transformation of scientific and technological achievements in colleges and universities across the country, while rarely divide the research objects in detail. In order to improve the evaluation effect of scientific and technological achievements in colleges and universities, this paper builds a university science and technology achievement evaluation system based on machine learning and image feature retrieval on the basis of analyzing the needs of high-tech achievement evaluation. The system has certain flexibility. Moreover, this study selects the appropriate network architecture based on the actual data and mission objectives of the high-tech achievement evaluation. In addition, this paper proposes a FT-GRU model of a gated recurrent unit network incorporating N nearest neighbor text, and a more stable model structure is obtained through system optimization. Finally, this study designs experiments to verify the performance of the model. The research results show that the university science and technology achievement evaluation system based on machine learning and image feature retrieval constructed in this study meets the expected goals and has certain practical significance.

Digitalization system of ancient architecture decoration art based on neural network and image features

The inheritance and innovation of ancient architecture decoration art is an important way for the development of the construction industry. The data process of traditional ancient architecture decoration art is relatively backward, which leads to the obvious distortion of the digitalization of ancient architecture decoration art. In order to improve the digital effect of ancient architecture decoration art, based on neural network, this paper combines the image features to construct a neural network-based ancient architecture decoration art data system model, and graphically expresses the static construction mode and dynamic construction process of the architecture group. Based on this, three-dimensional model reconstruction and scene simulation experiments of architecture groups are realized. In order to verify the performance effect of the system proposed in this paper, it is verified through simulation and performance testing, and data visualization is performed through statistical methods. The result of the study shows that the digitalization effect of the ancient architecture decoration art proposed in this paper is good.

An Adaptive Deep Belief Network With Sparse Restricted Boltzmann Machines

Deep belief network (DBN) is an efficient learning model for unknown data representation, especially nonlinear systems. However, it is extremely hard to design a satisfactory DBN with a robust structure because of traditional dense representation. In addition, backpropagation algorithm-based fine-tuning tends to yield poor performance since its ease of being trapped into local optima. In this article, we propose a novel DBN model based on adaptive sparse restricted Boltzmann machines (AS-RBM) and partial least square (PLS) regression fine-tuning, abbreviated as ARP-DBN, to obtain a more robust and accurate model than the existing ones. First, the adaptive learning step size is designed to accelerate an RBM training process, and two regularization terms are introduced into such a process to realize sparse representation. Second, initial weight derived from AS-RBM is further optimized via layer-by-layer PLS modeling starting from the output layer to input one. Third, we present the convergence and stability analysis of the proposed method. Finally, our approach is tested on Mackey-Glass time-series prediction, 2-D function approximation, and unknown system identification. Simulation results demonstrate that it has higher learning accuracy and faster learning speed. It can be used to build a more robust model than the existing ones.

A Multiscale Spatio-Temporal Convolutional Deep Belief Network for Sensor Fault Detection of Wind Turbine

Sensor fault detection of wind turbines plays an important role in improving the reliability and stable operation of turbines. The supervisory control and data acquisition (SCADA) system of a wind turbine provides promising insights into sensor fault detection due to the accessibility of the data and the abundance of sensor information. However, SCADA data are essentially multivariate time series with inherent spatio-temporal correlation characteristics, which has not been well considered in the existing wind turbine fault detection research. This paper proposes a novel classification-based fault detection method for wind turbine sensors. To better capture the spatio-temporal characteristics hidden in SCADA data, a multiscale spatio-temporal convolutional deep belief network (MSTCDBN) was developed to perform feature learning and classification to fulfill the sensor fault detection. A major superiority of the proposed method is that it can not only learn the spatial correlation information between several different variables but also capture the temporal characteristics of each variable. Furthermore, this method with multiscale learning capability can excavate interactive characteristics between variables at different scales of filters. A generic wind turbine benchmark model was used to evaluate the proposed approach. The comparative results demonstrate that the proposed method can significantly enhance the fault detection performance.

SVRG-MKL: A Fast and Scalable Multiple Kernel Learning Solution for Features Combination in Multi-Class Classification Problems

In this paper, we present a novel strategy to combine a set of compact descriptors to leverage an associated recognition task. We formulate the problem from a multiple kernel learning (MKL) perspective and solve it following a stochastic variance reduced gradient (SVRG) approach to address its scalability, currently an open issue. MKL models are ideal candidates to jointly learn the optimal combination of features along with its associated predictor. However, they are unable to scale beyond a dozen thousand of samples due to high computational and memory requirements, which severely limits their applicability. We propose SVRG-MKL, an MKL solution with inherent scalability properties that can optimally combine multiple descriptors involving millions of samples. Our solution takes place directly in the primal to avoid Gram matrices computation and memory allocation, whereas the optimization is performed with a proposed algorithm of linear complexity and hence computationally efficient. Our proposition builds upon recent progress in SVRG with the distinction that each kernel is treated differently during optimization, which results in a faster convergence than applying off-the-shelf SVRG into MKL. Extensive experimental validation conducted on several benchmarking data sets confirms a higher accuracy and a significant speedup of our solution. Our technique can be extended to other MKL problems, including visual search and transfer learning, as well as other formulations, such as group-sensitive (GMKL) and localized MKL (LMKL) in convex settings.

Early Fault Detection of Wind Turbines Based on Operational Condition Clustering and Optimized Deep Belief Network Modeling

Health monitoring and early fault detection of wind turbines have attracted considerable attention due to the benefits of improving reliability and reducing the operation and maintenance costs of the turbine. However, dynamic and constantly changing operating conditions of wind turbines still pose great challenges to effective and reliable fault detection. Most existing health monitoring approaches mainly focus on one single operating condition, so these methods cannot assess the health status of turbines accurately, leading to unsatisfactory detection performance. To this end, this paper proposes a novel general health monitoring framework for wind turbines based on supervisory control and data acquisition (SCADA) data. A key feature of the proposed framework is that it first partitions the turbine operation into multiple sub-operation conditions by the clustering approach and then builds a normal turbine behavior model for each sub-operation condition. For normal behavior modeling, an optimized deep belief network is proposed. This optimized modeling method can capture the sophisticated nonlinear correlations among different monitoring variables, which is helpful to enhance the prediction performance. A case study of main bearing fault detection using real SCADA data is used to validate the proposed approach, which demonstrates its effectiveness and advantages.