We address the problem of abnormal event detection from trajectory data. In this paper, a new adversarial approach is proposed for building a deep neural network binary classifier, trained in an unsupervised fashion, ...
详细信息
ISBN:
(纸本)9781728118383
We address the problem of abnormal event detection from trajectory data. In this paper, a new adversarial approach is proposed for building a deep neural network binary classifier, trained in an unsupervised fashion, that can distinguish normal from abnormal trajectory-based events without the need for setting manual detection threshold. Inspired by the generative adversarial network (GAN) framework, our GAN version is a discriminative one in which the discriminator is trained to distinguish normal and abnormal trajectory reconstruction errors given by a deep autoencoder. With urban traffic videos and their associated trajectories, our proposed method gives the best accuracy for abnormal trajectory detection. In addition, our model can easily be generalized for abnormal trajectory-based event detection and can still yield the best behavioural detection results as demonstrated on the CAVIAR dataset.
The brain-like functionality of the artificial neural networks besides their great performance in various areas of scientific applications, make them a reliable tool to be employed in Audio-Visual Speech Recognition (...
详细信息
The brain-like functionality of the artificial neural networks besides their great performance in various areas of scientific applications, make them a reliable tool to be employed in Audio-Visual Speech Recognition (AVSR) systems. The applications of such networks in the AVSR systems extend from the preliminary stage of feature extraction to the higher levels of information combination and speech modeling. In this paper, some carefully designed deep autoencoders are proposed to produce efficient bimodal features from the audio and visual stream inputs. The basic proposed structure is modified in three proceeding steps to make better usage of the presence of the visual information from the speakers' lips Region of Interest (ROI). The performance of the proposed structures is compared to both the unimodal and bimodal baselines in a professional phoneme recognition task, under different noisy audio conditions. This is done by employing a state-of-the-art DNN-HMM hybrid as the speech classifier. In comparison to the MFCC audio-only features, the finally proposed bimodal features cause an average relative reduction of 36.9% for a range of different noisy conditions, and also, a relative reduction of 19.2% for the clean condition in terms of the Phoneme Error Rates (PER). (C) 2018 Elsevier Inc. All rights reserved.
Heating, Ventilation, and Air Conditioning (HVAC) systems are generally built in a modular manner, comprising several identical subsystems in order to achieve their nominal capacity. These parallel subsystems and elem...
详细信息
ISBN:
(纸本)9783319651729;9783319651712
Heating, Ventilation, and Air Conditioning (HVAC) systems are generally built in a modular manner, comprising several identical subsystems in order to achieve their nominal capacity. These parallel subsystems and elements should have the same behavior and, therefore, differences between them can reveal failures and inefficiency in the system. The complexity in HVAC systems comes from the number of variables involved in these processes. For that reason, dimensionality reduction techniques can be a useful approach to reduce the complexity of the HVAC data and study their operation. However, for most of these techniques, it is not possible to project new data without retraining the projection and, as a result, it is not possible to easily compare several projections. In this paper, a method based on deep autoencoders is used to create a reference model with a HVAC system and new data is projected using this model to be able to compare them. The proposed approach is applied to real data from a chiller with 3 identical compressors at the Hospital of Leon.
In the field of an unmanned aerial vehicle (UAV), the navigation algorithm with high precision and easy implementation is a hot topic of research, and the key of UAV control is to obtain accurate and real-time attitud...
详细信息
ISBN:
(纸本)9781728103778
In the field of an unmanned aerial vehicle (UAV), the navigation algorithm with high precision and easy implementation is a hot topic of research, and the key of UAV control is to obtain accurate and real-time attitude information. In this paper, a feature fusion algorithm based on unsupervised deep autoencoder (DAE) is proposed. It is used for data fusion of multiple sensors. The experimental results show that the unsupervised feature fusion algorithm can effectively improve the accuracy and has the potential to be applied to the data fusion of UAV sensors.
This paper leverages heterogeneous auxiliary information to address the data sparsity problem of recommender systems. We propose a model that learns a shared feature space from heterogeneous data, such as item descrip...
详细信息
ISBN:
(纸本)9781538691595
This paper leverages heterogeneous auxiliary information to address the data sparsity problem of recommender systems. We propose a model that learns a shared feature space from heterogeneous data, such as item descriptions, product tags and online purchase history, to obtain better predictions. Our model consists of autoencoders, not only for numerical and categorical data, but also for sequential data, which enables capturing user tastes, item characteristics and the recent dynamics of user preference. We learn the autoencoder architecture for each data source independently in order to better model their statistical properties. Our evaluation on two MovieLens datasets and an e-commerce dataset shows that mean average precision and recall improve over state-of-the-art methods.
Learning graph representations generally indicate mapping the vertices of a graph into a low-dimension space, in which the proximity of the original data can be preserved in the latent space. However, traditional meth...
详细信息
ISBN:
(纸本)9783319937137;9783319937120
Learning graph representations generally indicate mapping the vertices of a graph into a low-dimension space, in which the proximity of the original data can be preserved in the latent space. However, traditional methods that based on adjacent matrix suffered from high computational cost when encountering large graphs. In this paper, we propose a deep autoencoder driven streaming methods to learn low-dimensional representations for graphs. The proposed method process the graph as a data stream fulfilled by sampling strategy to avoid straight computation over the large adjacent matrix. Moreover, a graph regularized deep autoencoder is employed in the model to keep different aspects of proximity information. The regularized framework is able to improve the representation power of learned features during the learning process. We evaluate our method in clustering task by the features learned from our model. Experiments show that the proposed method achieves competitive results comparing with methods that directly apply deep models over the complete graphs.
This paper presents a novel approach to detect changes in satellite images taken from the same location at different timestamps. Different change detection methods are applied to multispectral satellite images taken w...
详细信息
ISBN:
(纸本)9781538676936
This paper presents a novel approach to detect changes in satellite images taken from the same location at different timestamps. Different change detection methods are applied to multispectral satellite images taken with the Worldview-2 (WV-2) satellite, as well as to several of their feature indices such as normalized difference vegetation index (NDVI), normalized difference soil index (NDSI), non-homogeneous feature index (NHFD) and red-blue ratio (R/B). Besides, an additional image is used to remove temporary changes like vehicles, persons etc. The combination of changes is computed with a set of pixel-wise operations, and morphological filters are applied to improve the final change map. The combination of the satellite images with their feature indices proved to produce better results than computing the changes independently. This paper summarizes the methodology and presents the results obtained.
Massive and dynamic networks arise in many practical applications such as social media, security and public health. Given an evolutionary network, it is crucial to detect structural anomalies, such as vertices and edg...
详细信息
ISBN:
(纸本)9781450355520
Massive and dynamic networks arise in many practical applications such as social media, security and public health. Given an evolutionary network, it is crucial to detect structural anomalies, such as vertices and edges whose "behaviors" deviate from underlying majority of the network, in a real-time fashion. Recently, network embedding has proven a powerful tool in learning the low-dimensional representations of vertices in networks that can capture and preserve the network structure. However, most existing network embedding approaches are designed for static networks, and thus may not be perfectly suited for a dynamic environment in which the network representation has to be constantly updated. In this paper, we propose a novel approach, NETWALK, for anomaly detection in dynamic networks by learning network representations which can be updated dynamically as the network evolves. We first encode the vertices of the dynamic network to vector representations by clique embedding, which jointly minimizes the pairwise distance of vertex representations of each walk derived from the dynamic networks, and the deep autoencoder reconstruction error serving as a global regularization. The vector representations can be computed with constant space requirements using reservoir sampling. On the basis of the learned low-dimensional vertex representations, a clustering-based technique is employed to incrementally and dynamically detect network anomalies. Compared with existing approaches, NETWALK has several advantages: 1) the network embedding can be updated dynamically, 2) streaming network nodes and edges can be encoded efficiently with constant memory space usage, 3). flexible to be applied on different types of networks, and 4) network anomalies can be detected in real-time. Extensive experiments on four real datasets demonstrate the effectiveness of NETWALK.
The paper is devoted to the quantization algorithm development based on the neural networks framework. This research is considered in the context of the scalable real-time audio/speech coder based on the perceptually ...
详细信息
ISBN:
(纸本)9783319925370;9783319925363
The paper is devoted to the quantization algorithm development based on the neural networks framework. This research is considered in the context of the scalable real-time audio/speech coder based on the perceptually adaptive matching pursuit algorithm. The encoder parameterizes the input sound signal frame with some amount of real numbers that are need to be compactly represented in binary form, i.e. quantized. The neural network quantization approach gives great opportunity for such a goal because the data quantized in whole vector but not in separate form and it can effectively use correlations between each element of the input coded vector. deep autoencoder (DAE) neural network-based architecture for the quantization part of the encoding algorithm is shown. Its structure and learning features are described. Conducted experiments points out the big compression ratio with high reconstructed signal quality of the developed audio/speech coder quantization scheme.
Digital transformation of the world goes very fast during last two decades. Today, data is power and very important. Firstly, magnetic tapes and then digital data storages have been used to collect all data. After thi...
详细信息
ISBN:
(纸本)9781538641842
Digital transformation of the world goes very fast during last two decades. Today, data is power and very important. Firstly, magnetic tapes and then digital data storages have been used to collect all data. After this process, big data and its tool machine learning became very popular in both literature and industry. People use machine learning in order to obtain meaningful information from the big data. It brings valuable planning results. However, nowadays it is quite hard to collect and store all digital data to computers. This process is expensive and we will not have enough space to store data in the future. Therefore, we need and propose "Digital Data Forgetting" phrase with machine learning approach. With this digital / software solution, we will have more valuable data and will be able to erase the rest of them. We called this operation "Big Cleaning". In this article, we use a data set to get and extract more valuable data with principal component analysis (PCA), deep autoencoder and k-nearest neighbor machine learning methods in the experimental analysis section.
暂无评论