Classification, target detection, and compression are all important tasks in analyzing hyperspectral imagery (HSI). Because of the high dimensionality of HSI, it is often useful to identify low-dimensional representat...
详细信息
Classification, target detection, and compression are all important tasks in analyzing hyperspectral imagery (HSI). Because of the high dimensionality of HSI, it is often useful to identify low-dimensional representations of HSI data that can be used to make analysis tasks tractable. Traditional linear dimensionality reduction (DR) methods are not adequate due to the nonlinear distribution of HSI data. Many nonlinear DR methods, which are successful in the general data processing domain, such as Local Linear Embedding (LLE) [1], Isometric Feature Mapping (ISOMAP) [2] and Kernel Principal Components Analysis (KPCA) [3], run very slowly and require large amounts of of memory when applied to HSI. For example, applying KPCA to the 512×217 pixel, 204-band Salinas image using a modern desktop computer (AMD FX-6300 Six-Core Processor, 32 GB memory) requires more than 5 days of computing time and 28GB memory! In this thesis, we propose two different algorithms for significantly improving the computational efficiency of nonlinear DR without adversely affecting the performance of classification task: Simple Linear Iterative Clustering (SLIC) superpixels and semi-supervised deep autoencoder networks (SSDAN). SLIC is a very popular algorithm developed for computing superpixels in RGB images that can easily be extended to HSI. Each superpixel includes hundreds or thousands of pixels based on spatial and spectral similarities and is represented by the mean spectrum and spatial position of all of its component pixels. Since the number of superpixels is much smaller than the number of pixels in the image, they can be used as input for nonlinearDR, which significantly reduces the required computation time and memory versus providing all of the original pixels as input. After nonlinear DR is performed using superpixels as input, an interpolation step can be used to obtain the embedding of each original image pixel in the low dimensional space. To illustrate the power of using superpi
Explainable neural models have gained a lot of attention in recent years. However, conventional encoder–decoder models do not capture information regarding the importance of the involved latent variables and rely on ...
详细信息
Explainable neural models have gained a lot of attention in recent years. However, conventional encoder–decoder models do not capture information regarding the importance of the involved latent variables and rely on a heuristic a-priori specification of the dimensionality of the latent space or its selection based on multiple trainings. In this paper, we focus on the efficient structuring of the latent space of encoder–decoder approaches for explainable data reconstruction and compression. For this purpose, we leverage the concept of Shapley values to determine the contribution of the latent variables on the model’s output and rank them according to decreasing importance. As a result, a truncation of the latent dimensions to those that contribute the most to the overall reconstruction allows a trade-off between model compactness (i.e. dimensionality of the latent space) and representational power (i.e. reconstruction quality). In contrast to other recent autoencoder variants that incorporate a PCA-based ordering of the latent variables, our approach does not require time-consuming training processes and does not introduce additional weights. This makes our approach particularly valuable for compact representation and compression. We validate our approach at the examples of representing and compressing images as well as high-dimensional reflectance data.
Predictive maintenance uses improved Fuzzy Analytic Hierarchy Process(FAHP) model to calculate the weight of equipment failure and the comprehensive index information of equipment failure, to identify its deterioratio...
详细信息
Predictive maintenance uses improved Fuzzy Analytic Hierarchy Process(FAHP) model to calculate the weight of equipment failure and the comprehensive index information of equipment failure, to identify its deterioration trend, and to judge its trend, to calculate the comprehensive maintenance threshold, to generate maintenance decision information and to identify the equipment locations that need to be disposed of.
This paper proposes a modulation classification method based on stacked denoising autoencoders (SDAE). This method can extract the modulation features automatically and classify the input signals based on the extracte...
详细信息
This paper proposes a modulation classification method based on stacked denoising autoencoders (SDAE). This method can extract the modulation features automatically and classify the input signals based on the extracted features. The scenarios of rapid classification and high-accuracy classification are considered. In a rapid classification scenario, the classification speed has priority over the classification accuracy. Therefore, a long-symbol sequence is not attainable for this scenario. Moreover, expert features are not necessary for this scenario, simplifying the modulation classification procedure and rendering rapid classification more achievable. In addition, in a high-accuracy classification scenario, higher cumulants are used as the expert features owing to their advantage over the other features at noise resistance. We use complex symbols rather than pulse shaped complex signals as the network inputs, simplifying the network topology and reducing the calculation overhead. The results of the average classification accuracy, the individual classification accuracy, the execution time and the influence of the signal sampling synchronization are presented, demonstrating significant performance advantages over the other methods. Copyright (c) 2016 John Wiley & Sons, Ltd.
Hyperspectral images contain very useful information because of their high spectral resolution, which can be used for non-contact food testing. However, to process them with convolutional neural networks, large data s...
详细信息
Hyperspectral images contain very useful information because of their high spectral resolution, which can be used for non-contact food testing. However, to process them with convolutional neural networks, large data sets are needed. This is especially true if the data is not preprocessed and therefore of high dimension. However, relatively few hyperspectral data sets exist. To solve this problem, the neural network can be pre-trained using an autoencoder, which compresses and reconstructs the image. By minimizing the reconstruction error, useful features can be learned to solve the original task. In this work, spice mixtures are used to investigate whether individual components can be detected. In particular, a neural network using a 3D convolutional autoencoder is trained with a small data set.
Tens of millions of people live blind, and their number is ever increasing. Visual-to-auditory sensory substitution (SS) encompasses a family of cheap, generic solutions to assist the visually impaired by conveying vi...
详细信息
Tens of millions of people live blind, and their number is ever increasing. Visual-to-auditory sensory substitution (SS) encompasses a family of cheap, generic solutions to assist the visually impaired by conveying visual information through sound. The required SS training is lengthy: months of effort is necessary to reach a practical level of adaptation. There are two reasons for the tedious training process: the elongated substituting audio signal, and the disregard for the compressive characteristics of the human hearing system.
To overcome these obstacles, we developed a novel class of SS methods, by training deep recurrent autoencoders for image-to-sound conversion. We successfully trained deep learning models on different datasets to execute visual-to-auditory stimulus conversion. By constraining the visual space, we demonstrated the viability of shortened substituting audio signals, while proposing mechanisms, such as the integration of computational hearing models, to optimally convey visual features in the substituting stimulus as perceptually discernible auditory components. We tested our approach in two separate cases. In the first experiment, the author went blindfolded for 5 days, while performing SS training on hand posture discrimination. The second experiment assessed the accuracy of reaching movements towards objects on a table. In both test cases, above-chance-level accuracy was attained after a few hours of training.
Our novel SS architecture broadens the horizon of rehabilitation methods engineered for the visually impaired. Further improvements on the proposed model shall yield hastened rehabilitation of the blind and a wider adaptation of SS devices as a consequence.
In this paper, we describe a intrusion detection algorithm based on deep learning for industrial control networks, aiming at the security problem of industrial control network. Deep learning is a kind of intelligent a...
详细信息
ISBN:
(纸本)9781450376228
In this paper, we describe a intrusion detection algorithm based on deep learning for industrial control networks, aiming at the security problem of industrial control network. Deep learning is a kind of intelligent algorithm and has the ability of automatically learning. It use self-learning to enhance the experience and dynamic classification capabilities. The ideology of deep learning is similar to the idea of intrusion detection to improve the detection rate and reduce the rate of false through learning, a sparse auto-encoder-extreme learning machine intrusion detection model is proposed for the problem of intrusion detection accuracy. It uses deep learning autoencoder to combine the coefficient penalty and reconstruction loss of the encode layer to extract the features of high-dimensional data during the training model, and then uses the extreme learning machine to quickly and effectively classify the extracted features. The accuracy of the algorithm is verified by the industrial control intrusion detection standard data set. The experimental results verify that the method can effectively improve the performance of the intrusion detection system and reduce the false alarm rate.
Breast cancer is the most common cause of cancer death in women. Today, post-transcriptional protein products of the genes involved in breast cancer can be identified by immunohistochemistry. However, this method has ...
详细信息
Breast cancer is the most common cause of cancer death in women. Today, post-transcriptional protein products of the genes involved in breast cancer can be identified by immunohistochemistry. However, this method has problems arising from the intra-observer and inter-observer variability in the assessment of pathologic variables, which may result in misleading conclusions. Using an optimal selection of preprocessing techniques may help to reduce observer variability. Deep learning has emerged as a powerful technique for any tasks related to machine learning such as classification and regression. The aim of this work is to use autoencoders (neural networks commonly used to feed deep learning architectures) to improve the quality of the data for developing immunohistochemistry signatures with prognostic value in breast cancer. Our testing on data from 222 patients with invasive non-special type breast carcinoma shows that an automatic binarization of experimental data after autoencoding could outperform other classical preprocessing techniques (such as human-dependent or automatic binarization only) when applied to the prognosis of breast cancer by immunohistochemical signatures. (C) 2017 Elsevier Inc. All rights reserved.
As one of the most rapidly developing artificial intelligence techniques, deep learning has been applied in various machine learning tasks and has received great attention in data science and statistics. Regardless of...
详细信息
As one of the most rapidly developing artificial intelligence techniques, deep learning has been applied in various machine learning tasks and has received great attention in data science and statistics. Regardless of the complex model structure, deep neural networks can be viewed as a nonlinear and nonparametric generalization of existing statistical models. In this review, we introduce several popular deep learning models including convolutional neural networks, generative adversarial networks, recurrent neural networks, and autoencoders, with their applications in image data, sequential data and recommender systems. We review the architecture of each model and highlight their connections and differences compared with conventional statistical models. In particular, we provide a brief survey of the recent works on the unique overparameterization phenomenon, which explains the strengths and advantages of using an extremely large number of parameters in deep learning. In addition, we provide a practical guidance on optimization algorithms, hyperparameter tuning, and computing resources.
A fall is an abnormal activity that occurs rarely, so it is hard to collect real data for falls. It is, therefore, difficult to use supervised learning methods to automatically detect falls. Another challenge in autom...
详细信息
A fall is an abnormal activity that occurs rarely, so it is hard to collect real data for falls. It is, therefore, difficult to use supervised learning methods to automatically detect falls. Another challenge in automatically detecting falls is the choice of engineered features. In this paper, we formulate fall detection as an anomaly detection problem and propose to use an ensemble of autoencoders to learn features from different channels of wearable sensor data trained only on normal activities. We show that the traditional approach of choosing a threshold as the maximum of the reconstruction error on the training normal data is not the right way to identify unseen falls. We propose two methods for automatic tightening of reconstruction error from only the normal activities for better identification of unseen falls. We present our results on two activity recognition datasets and show the efficacy of our proposed method against traditional autoencoder models and two standard one-class classification methods. (c) 2017 Elsevier Ltd. All rights reserved.
暂无评论