Attention of drivers is very important for road safety and it is worth observing even in laboratory conditions during a simulated drive. This paper deals with design of an experiment investigating driver's attenti...
详细信息
ISBN:
(纸本)9789897582134
Attention of drivers is very important for road safety and it is worth observing even in laboratory conditions during a simulated drive. This paper deals with design of an experiment investigating driver's attention, validation of collected data, and first preprocessing and processing steps used within data analysis. Brain activity is considered as a primary biosignal and is measured and analyzed using the techniques and methods of electroencephalography and event related potentials. Respiration is considered as a secondary biosignal that is captured together with brain activity. Validation of collected data using a stacked autoencoder is emphasized as an important step preceding data analysis.
In wastewater treatment plants (WWTPs), some variables such as BOD5 and COD that are related to effuent quality, are difficult to measure directly online due to technical or economic limitations. To deal with this pro...
详细信息
In wastewater treatment plants (WWTPs), some variables such as BOD5 and COD that are related to effuent quality, are difficult to measure directly online due to technical or economic limitations. To deal with this problem, a soft sensor that is based on a deep neural network with a named stacked autoencoder (SAE) is developed for WWTPs. Neural networks with deep structure are superior to shallow ones when facing complex problems in modern applications, which makes them suitable for wastewater treatment processes. However, deep structures are difficult to train when using traditional learning algorithms, and there are no general guidelines for identifying the proper network structure for a specific application. In the present work, a deep learning technique called the greedy layer-wise training algorithm is employed to train a deep neural network, and a genetic-algorithm strategy is developed for identifying the number of neurons in each hidden layer. In order to demonstrate its usefulness, the proposed soft sensor is tested through the test-bed Benchmark Simulation Model No. 1 (BSM1) with different weather conditions. The results indicate that the proposed soft sensor based on a deep-structure neural network can achieve better prediction and generalization performance in comparison with commonly used methodologies.
Many classification problems encountered in real-world applications exhibit a profile of imbalanced data. Current methods depend on data resampling. In fact, if the feature set provides a clear decision boundary, resa...
详细信息
Many classification problems encountered in real-world applications exhibit a profile of imbalanced data. Current methods depend on data resampling. In fact, if the feature set provides a clear decision boundary, resampling may not be needed to solve the imbalanced classification problem. Therefore, this work proposes a feature learning method based on the autoencoder to learn a set of features with better classification capabilities of the minority and the majority classes to address the imbalanced classification problems. Two sets of features are learned by two stacked autoencoders with different activation functions to capture different characteristics of the data and they are combined to form the Dual Autoencoding Features. Samples are then classified in the new feature space learned in this manner instead of the original input space. Experimental results show that the proposed method outperforms current resampling-based methods with statistical significance for imbalanced pattern classification problems. (C) 2016 Elsevier Ltd. All rights reserved.
The emergence of mobile health (mHealth) systems has risen the challenges and concerns due to the sensitivity of the data involved in such systems. It is essential to ensure that these data are well delivered to the h...
详细信息
ISBN:
(纸本)9781509043729
The emergence of mobile health (mHealth) systems has risen the challenges and concerns due to the sensitivity of the data involved in such systems. It is essential to ensure that these data are well delivered to the health monitoring center for accurate and perfect diagnosis and follow-up. Due to the wireless network constraints, these requirements become more challenging. In this paper, we propose a deep learning approach for EEG data compression in mHealth system. We show that the stacked autoencoder neural network architecture is efficient for EEG data compression. We conduct a comprehensive comparative study that demonstrates the effectiveness of our system for EEG compression in addition to preserving the total energy consumption.
Functional connectivity (FC) patterns obtained from resting-state functional magnetic resonance imaging data are commonly employed to study neuropsychiatric conditions by using pattern classifiers such as the support ...
详细信息
Functional connectivity (FC) patterns obtained from resting-state functional magnetic resonance imaging data are commonly employed to study neuropsychiatric conditions by using pattern classifiers such as the support vector machine (SVM). Meanwhile, a deep neural network (DNN) with multiple hidden layers has shown its ability to systematically extract lower-to-higher level information of image and speech data from lower-to-higher hidden layers, markedly enhancing classification accuracy. The objective of this study was to adopt the DNN for whole-brain resting-state FC pattern classification of schizophrenia (SZ) patients vs. healthy controls (HCs) and identification of aberrant FC patterns associated with SZ. We hypothesized that the lower-to-higher level features learned via the DNN would significantly enhance the classification accuracy, and proposed an adaptive learning algorithm to explicitly control the weight sparsity in each hidden layer via L1-norm regularization. Furthermore, the weights were initialized via stacked autoencoder based pre-training to further improve the classification performance. Classification accuracy was systematically evaluated as a function of (1) the number of hidden layers/nodes, (2) the use of L1-norm regularization, (3) the use of the pre-training, (4) the use of frame wise displacement (FD) removal, and (5) the use of anatomical/functional parcellation. Using FC patterns from anatomically parcellated regions without FD removal, an error rate of 14.2% was achieved by employing three hidden layers and 50 hidden nodes with both L1-norm regularization and pre-training, which was substantially lower than the error rate from the SVM (22.3%). Moreover, the trained DNN weights (i.e., the learned features) were found to represent the hierarchical organization of aberrant FC patterns in SZ compared with HC. Specifically, pairs of nodes extracted from the lower hidden layer represented sparse FC patterns implicated in SZ, which was quantifie
We propose a novel Deep learning approach using autoencoders to map spectral bands to a space of lower dimensionality while preserving the information that makes it possible to discriminate different materials. Deep l...
详细信息
ISBN:
(纸本)9781510603806;9781510603813
We propose a novel Deep learning approach using autoencoders to map spectral bands to a space of lower dimensionality while preserving the information that makes it possible to discriminate different materials. Deep learning is a relatively new pattern recognition approach which has given promising result in many applications. In Deep learning a hierarchical representation of increasing level of abstraction of the features is learned. autoencoder is an important unsupervised technique frequently used in Deep learning for extracting important properties of the data. The learned latent representation is a non-linear mapping of the original data which potentially preserve the discrimination capacity.
Personal and easy-to-use health checking system is an attractive application of sensor systems. Sensing data analysis for diagnosis is important as well as preparing small and mobile sensor nodes because sensing data ...
详细信息
ISBN:
(纸本)9781509032198
Personal and easy-to-use health checking system is an attractive application of sensor systems. Sensing data analysis for diagnosis is important as well as preparing small and mobile sensor nodes because sensing data include variations and noises reflecting individual difference of people and sensing conditions. Deep Neural Network, or Deep Learning, is a well-known method of machine learning and it is effective for feature extraction from pictures. Then, we thought Deep Learning also can extract features from sensing data. In this paper, we tried to build a diagnosis system of lung cancer based on Deep Learning. Input data of the system was generated from human urine by Gas Chromatography Mass Spectrometer (GC-MS) and our system achieved 90% accuracy in judging whether the patient had lung cancer or not. This system will be useful for pre- and personal diagnosis because collecting urine is very easy and not harmful to human body. We are targeting installation of this system not only to gas chromatography systems but also to some combination of multiple sensors for detecting gases of low concentration.
The work presents an approach towards facial emotion recognition using face dataset consisting of four classes of emotions (happy, angry, neutral and sad) with different models of deep neural networks and compares the...
详细信息
ISBN:
(纸本)9781467399104
The work presents an approach towards facial emotion recognition using face dataset consisting of four classes of emotions (happy, angry, neutral and sad) with different models of deep neural networks and compares their performance. We take the raw pixels values of all images in CMU face images dataset. The pixels values were represented by higher level concepts by feeding them into Restricted Boltzmann Machine, Deep Belief Networks and stacked autoencoder with Softmax Function. We observe that the later model could learn to recognize emotion with significantly higher accuracy compared to the former two models. Also, its performance improves with an increase in the number of hidden nodes in autoencoders, unlike the other two models.
Motivated by deep learning approaches to classify normal and neuro-diseased subjects in functional Magnetic Resonance Imaging (fMRI), we propose stacked autoencoder (SAE) based 2-stage architecture for disease diagnos...
详细信息
ISBN:
(纸本)9781450347532
Motivated by deep learning approaches to classify normal and neuro-diseased subjects in functional Magnetic Resonance Imaging (fMRI), we propose stacked autoencoder (SAE) based 2-stage architecture for disease diagnosis. In the proposed architecture, a separate 4-hidden layer autoencoder is trained in unsupervised manner for feature extraction corresponding to every brain region. Thereafter, these trained autoencoders are used to provide features on class-labeled input data for training a binary support vector machine (SVM) based classifier. In order to design a robust classifier, noisy or inactive gray matter voxels are filtered out using a proposed covariance based approach. We applied the proposed methodology on a public dataset, namely, 1000 Functional Connectomes Project Cobre dataset consisting of fMRI data of normal and Schizophrenia subjects. The proposed architecture is able to classify normal and Schizophrenia subjects with 10-fold cross-validation accuracy of 92% that is better compared to the existing methods used on the same dataset.
Underwater target recognition remains a challenging task due to the complex and changeable environment. There have been a huge number of methods to deal with this problem. However, most of them fail to hierarchically ...
详细信息
ISBN:
(纸本)9781509041657
Underwater target recognition remains a challenging task due to the complex and changeable environment. There have been a huge number of methods to deal with this problem. However, most of them fail to hierarchically extract deep features. In this paper, a novel deep learning framework for underwater target classification is proposed. First, instead of extracting features relying on expert knowledge, sparse autoencoder (AE) is utilized to learn invariant features from the spectral data of underwater targets. Second, stacked autoencoder (SAE) is used to get high-level features as a deep learning method. At last, the joint of SAE and softmax is proposed to classify the underwater targets. Experiment results with the received signal data from three different targets on the sea indicated that the proposed approach can get the highest classification accuracy compared with support vector machine (SVM) and probabilistic neural network (PNN).
暂无评论