Fast Road Detection Methods on a Large Scale Dataset for Assisting Robot Navigation Using Kernel Principal Component Analysis and Deep Learning

Tire defects are crucial for safe driving. Specialized experts or expensive tools such as stereo depth cameras and depth gages are usually used to investigate these defects. In image processing, feature extraction, reduction, and classification are presented as three challenging and symmetric ways to affect the performance of machine learning models. This paper proposes a hybrid system for cracked tire detection based on the adaptive selection of correlation features and deep belief neural networks. The proposed system has three steps: feature extraction, selection, and classification. First, the oriented gradient histogram extracts features from the tire images. Second, the proposed adaptive correlation feature selection selects important features with a threshold value adapted to the nature of the images. The last step of the system is to predict the image category based on the deep belief neural networks technique. The proposed model is tested and evaluated using real images of cracked and normal tires. The experimental results show that the proposed solution performs better than the current studies in effectively classifying tire defect images. The proposed hybrid cracked tire detection system based on adaptive correlation feature selection and Deep Belief Neural Networks’ performance provided better classification accuracy (88.90%) than that of Belief Neural Networks (81.6%) and Convolution Neural Networks (85.59%).

Short-term load forecasting is very important for power system dispatching since the economy and stability of power system are affected by the accuracy of load forecasting. Therefore, a short-term load forecasting method based on deep belief neural network (DBN) is proposed in this paper. Firstly, the low-voltage user load data are normalized to construct the training set and test set for low-voltage user short-term load forecasting; then, the DBN model is trained and adjusted by using the training set to obtain the well-trained DBN network; finally, the historical data of short-term load of users are input for load forecasting. The actual load data in Zhejiang power grid of China are serviced for demonstrating the effectiveness the proposed algorithm. The simulation results show that the short-term load forecasting algorithm proposed in this paper is with high accuracy for the actual load forecasting of low voltage users.

In order to ensure the safe and stable operation of power system, enrich the means of power grid analysis and control, and expand the application of deep intelligent learning methods in power grid systems, the application of deep learning intelligent machine learning method in frequency prediction of large power grid is explored. First, on the basis of deep learning, the frequency response mode of large power grid is analyzed and the key characteristic quantities that affect the frequency response mode are extracted. Second, the deep belief neural network (DBN. DNN) frequency prediction model is constructed. Also, the training and testing of the model are introduced. Finally, the input and output based on the DBN.CNN prediction model and the network structure design of the model are analyzed. The prediction performance of the model is evaluated. The results show that when the number of neurons in the hidden layer is 50, the model achieves the optimal prediction effect. Increasing the number of training samples helps to improve the modeling ability and prediction accuracy of the model. For frequency prediction problems, the number of training samples should be set to ≥400, and the number of hidden layers corresponding to the model should be 5. When the number of hidden layer neurons is 10, the prediction accuracy of the DBN/DNN network is poor. When the number of hidden layer neurons is 50, the model can achieve the best prediction effect. Overall, the DBN.DNN prediction model has good prediction performance. The RMSE of the forecast data is O. 0073Hz can basically meet the actual application requirements. Therefore, the frequency prediction method based on deep belief neural network has certain advantages in accuracy and efficiency.

Due to the high probability of power flow violation during the N-1 static safety check under maintenance, and the risk assessment indicators are single and subjective, a method of static security risk assessment based on multi-source heterogeneous information for planned maintenance is proposed. By regularizing dispatcher experience, the static security factors are analyzed from equipment failure rate, electrical characteristics and grid topology, etc. Then, the selected indicators are preprocessed. Finally, Deep Belief Neural Network (DBNN) is used to evaluate risk. Through deep mining of the membership and mapping relationship between multiple indicators, the risk self-assessment is realized, and dispatcher is given the assistant decision when orderly adjusting equipment that causes limit. Simulation results of IEEE 39-bus show that the method proposed in this paper can quickly assess risk level without analysis of dispatcher; compared with DNN algorithm, DBNN has higher accuracy. The validity and feasibility of this research are proved.

1 - Congruence of deep learning in biomedical engineering: future prospects and challenges

Sketch based image retrieval using grid approach on large scale database

Lean based cycle time reduction in manufacturing companies using black widow based deep belief neural network

This paper proposes a new technique for the identification of fundamental heart sounds (HSs), namely, $S_{1}$ and $S_{2}$ from the cardiac cycle, the first and foremost step in the automated HSs analysis for the detection of pathological events, without incorporating time-interval informations between $S_{1}$ and $S_{2}$ or electrocardiogram signal as reference. The motive of this paper is to demonstrate that the reliable $S_{1}$ and $S_{2}$ classification performances based on the combinatory feature (CF) derived from higher order moments and cepstral-based domain can still be achieved, under the circumstances where the timing interval information might not be easily understood due to cardiac abnormalities. Using deep neural networks approach, a stacked autoencoder (SAE) based on the CF is proposed for the classification of fundamental HSs. Experiments are conducted on both publicly available and recorded HSs signals for the validation of the proposed method. The SAE using the proposed CF achieves better classification results in recognizing $S_{1}$ and $S_{2}$ in comparison to well-known classifiers such as deep belief neural network, support vector machine, Naive Bayes, linear discriminant analysis, and boosting ensemble. The proposed method shows higher classification rate in terms of accuracy, sensitivity, and specificity by considering CF, which uses Mel-frequency cepstral coefficients and its derivative features. A second approach for addressing the problem of $S_{1}$ and $S_{2}$ identifications is carried out by employing 1-D convolutional neural network that uses the signals directly to learn the relevant features by its own for the recognition.

Electroencephalography (EEG) is the oldest diagnostic tool used in the field of neurosciences. When a neurologist sees an EEG report, he can point out important neural defects in a person but many a times diagnoses were missed and it is just impossible for human brain to process all the data in EEG. Nowadays many deep machine learning architectures are developed to understand the information contained in EEG signals. This study reviews various literatures on classification of EEG signals using various artificial neural networks like Convoluted neural network, Recurrent Neural Network, Deep Belief neural networks and hybrids.

Wireless sensor networks (WSNs) have recently acquired prominence in a variety of applications such as remote monitoring and tracking. Since it is virtually hard to recharge the nodes in their remote deployment, also, the transmission of data from nodes to the base station requires a significant amount of energy. Thus, our research proposes a routing protocol, namely hybrid falcon-improved ACO Nature-Inspired Optimization using a deep learning model to reduce energy consumption while increases the network lifetime. In the developed model, initially, the falcon optimization technique is utilized to locate the best possible cluster heads in the quickest possible time. Furthermore, to improve the quality of service in routing optimization a new improved ACO has been proposed in which linear flexible operator and the premier operator are used to increasing the iteration speed. Finally, the optimum route is obtained through DBNN based on predicted energy. As a result, our proposed model gives a lifetime as 121 s and energy consumption as 0.041 J at 500 rounds when compared to the baseline approaches. Therefore, our proposed approaches provides better routing and improves the QoS as well as the energy consumption which increases the longevity of mobile nodes.

Recently there has been increasing interest in ways of using outof-domain (OOD) data to improve automatic speech recognition performance in domains where only limited data is available. This paper focuses on one such domain, namely that of disordered speech for which only very small databases exist, but where normal speech can be considered OOD. Standard approaches for handling small data domains use adaptation from OOD models into the target domain, but here we investigate an alternative approach with its focus on the feature extraction stage: OOD data is used to train feature-generating deep belief neural networks. Using AMI meeting and TED talk datasets, we investigate various tandem-based speaker independent systems as well as maximum a posteriori adapted speaker dependent systems. Results on the UAspeech isolated word task of disordered speech are very promising with our overall best system (using a combination of AMI and TED data) giving a correctness of 62.5%; an increase of 15% on previously best published results based on conventional model adaptation. We show that the relative benefit of using OOD data varies considerably from speaker to speaker and is only loosely correlated with the severity of a speaker’s impairments. Index Terms: Speech recognition, Tandem features, Deep belief neural network, Disordered speech

We study the eigenvalue distribution of random matrices pertinent to the analysis of deep neural networks. The matrices resemble the product of the sample covariance matrices, however, an important difference is that the analog of the population covariance matrix is now a function of random data matrices (synaptic weight matrices in the deep neural network terminology). The problem has been treated in recent work [J. Pennington, S. Schoenholz and S. Ganguli, The emergence of spectral universality in deep networks, Proc. Mach. Learn. Res. 84 (2018) 1924–1932, arXiv:1802.09979] by using the techniques of free probability theory. Since, however, free probability theory deals with population covariance matrices which are independent of the data matrices, its applicability in this case has to be justified. The justification has been given in [L. Pastur, On random matrices arising in deep neural networks: Gaussian case, Pure Appl. Funct. Anal. (2020), in press, arXiv:2001.06188] for Gaussian data matrices with independent entries, a standard analytical model of free probability, by using a version of the techniques of random matrix theory. In this paper, we use another version of the techniques to extend the results of [L. Pastur, On random matrices arising in deep neural networks: Gaussian case, Pure Appl. Funct. Anal. (2020), in press, arXiv:2001.06188] to the case where the entries of the data matrices are just independent identically distributed random variables with zero mean and finite fourth moment. This, in particular, justifies the mean field approximation in the infinite width limit for the deep untrained neural networks and the property of the macroscopic universality of random matrix theory in this case.

Bayesian Linear Regression analysis of the effect of silica fume on concrete creep predictions and structures’ safety

A large amount of video data is generated every day. Searching through huge video database is an important problem in many applications. For instance, individuals may want to search for video content they are interested in from YouTube video, media companies may want to locate video content that violates their copyright protection and so on. Fast and accurate algorithm in all these cases is needed for efficient video retrieval. The high dimensionality of video sequence poses a major challenge of video indexing and retrieval. As dimensionality increases, query performance degrades. This phenomenon generally referred to as the dimensionality curse, can be circumvented by reducing the dimensionality of the data. Such a reduction is however accompanied by loss of precision of query results. Feature extraction and dimensionality reduction are highly related to each other, as the combined goal of the two processing steps is to generate a compact representation of the content of an image. Here we propose to perform dimensionality reduction in both phases of the video search and retrieval system by extracting appropriate features. We shall try to exploit the use of principal component analysis to transform the original data of high dimensionality into new co-ordinate system with low dimensionality and then use sparse representation before applying similarity match for fast and accurate search and retrieval of videos.

Extracting and inserting knowledge into stacked denoising auto-encoders