Similarity (and distance metric) learning plays a very important role in many artificial intelligence tasks aiming at quantifying the relevance between objects. We address the challenge of learning complex relation pa...
详细信息
Similarity (and distance metric) learning plays a very important role in many artificial intelligence tasks aiming at quantifying the relevance between objects. We address the challenge of learning complex relation patterns from data objects exhibiting heterogeneous properties, and develop an effective multi view multimodal similarity learning model with much improved learning performance and model interpretability. The proposed method firstly computes multi-view convolutional features to achieve improved object representation, then analyses the similarities between objects by operating over multiple hidden relation types (modalities), and finally fine-tunes the entire model variables via back -propagating a ranking loss to the convolutional layers. We develop a topic-driven initialization scheme, so that each learned relation type can be interpreted as a representative of semantic topics of the objects. To improve model interpretability and generalization, sparsity is imposed over these hidden relations. The proposed method is evaluated by solving the image retrieval task using challenging image datasets, and is compared with seven state-of-the-art algorithms in the field. Experimental results demonstrate significant performance improvement of the proposed method over the competing ones. (C) 2017 Elsevier Ltd. All rights reserved.
In the past few years, a rapid increase in the number of patients requiring constant monitoring, which inspires researchers to develop intelligent and sustainable remote smart healthcare services. However, the transmi...
详细信息
ISBN:
(纸本)9781538677476
In the past few years, a rapid increase in the number of patients requiring constant monitoring, which inspires researchers to develop intelligent and sustainable remote smart healthcare services. However, the transmission of big real-time health data is a challenge since the current dynamic networks are limited by different aspects such as the bandwidth, end-to-end delay, and transmission energy. Due to this, a data reduction technique should be applied to the data before being transmitted based on the resources of the network. In this paper, we integrate efficient data reduction with wireless networking transmission to enable an adaptive compression with an acceptable distortion, while reacting to the wireless network dynamics such as channel fading and user mobility. convolutional auto-encoder (CAE) approach was used to implement an adaptive compression/reconstruction technique with the minimum distortion. Then, a resource allocation framework was implemented to minimize the transmission energy along with the distortion of the reconstructed signal while considering different network and applications constraints. A comparison between the results of the resource allocation framework considering both CAE and Discrete wavelet transforms (DWT) was also captured.
Objective: Abdominal ECG (AECG) recorded at the maternal abdomen is significantly affected by the maternal ECG (MECG), making the extraction of FECG a challenging task. This paper presents a new MECG elimination metho...
详细信息
ISBN:
(纸本)9781728118673
Objective: Abdominal ECG (AECG) recorded at the maternal abdomen is significantly affected by the maternal ECG (MECG), making the extraction of FECG a challenging task. This paper presents a new MECG elimination method based on Short Time Fourier Transform (STFT) and convolutionalautoencoder (CAE). Methods: First, the STFT is used to transform the AECG from one-dimensional (1D) time domain into two-dimensional(2D) time-frequency domain. Next, the CAE model is applied to estimate the 2D-STFT coefficients of MECG. Finally, after the inverse STFT of MECG, we can extract the FECG by subtracting the MECG from the AECG in the time domain. Different from the methods estimated the MECG in the 1D time domain, the novelty of the proposed method relies on estimating the MECG in the 2D time-frequency domain. Specifically, the CAE model learns the end-to-end mappings from the 2D-STFT coefficients of AECG to the MECG. Results: Experimental results show that the proposed method is effective in removing the MECG. Significance: This work enhances the clinical applications of FECG in the early detection of fetal heart diseases.
The accuracy and robustness of image classification with supervised deep learning are dependent on the availability of large-scale, annotated training data. However, there is a paucity of annotated data available due ...
详细信息
ISBN:
(纸本)9781538636411
The accuracy and robustness of image classification with supervised deep learning are dependent on the availability of large-scale, annotated training data. However, there is a paucity of annotated data available due to the complexity of manual annotation. To overcome this problem, a popular approach is to use transferable knowledge across different domains by: 1) using a generic feature extractor that has been pre-trained on large-scale general images (i.e., transfer learned) but which not suited to capture characteristics from medical images;or 2) fine-tuning generic knowledge with a relatively smaller number of annotated images. Our aim is to reduce the reliance on annotated training data by using a new hierarchical unsupervised feature extractor with a convolutional auto-encoder placed atop of a pre-trained convolutional neural network. Our approach constrains the rich and generic image features from the pre-trained domain to a sophisticated representation of the local image characteristics from the unannotated medical image domain. Our approach has a higher classification accuracy than transfer-learned approaches and is competitive with state-ofthe-art supervised fine-tuned methods.
The paper is motivated by the fact that brain cancer is one of the deadliest cancers and its detection in early stages is of paramount importance. In this regard, tumor 3D shape reconstruction from magnetic resonance ...
详细信息
ISBN:
(纸本)9781479970612
The paper is motivated by the fact that brain cancer is one of the deadliest cancers and its detection in early stages is of paramount importance. In this regard, tumor 3D shape reconstruction from magnetic resonance (MR) or computed tomography (CT) scans provides critical information, which can not be interpreted from 2D images. However, CT and MR images have low resolution in z direction compared to their resolution in x and y directions, therefore, 3D reconstructed shapes are of low quality. In this paper, we propose to use convolutional auto-encoders (CAEs) to address this drawback, and develop a convolutionalautoencoder-based inter-slice interpolation (CARISI) framework. Although deep nets have been used very recently for brain tumor segmentation, to the best of our knowledge, this is the first attempt to use CAEs for 3D reconstruction of brain tumor. The proposed CARISI framework consists of several encoding and decoding components, which can handle rapid changes in tumor shape without the need for supervision of an expert. Our experiments based on a real data-set consisting of 3064 segmented brain tumor images indicate that the proposed CARISI framework outperforms its counterpart and has the potential to significantly improve the overall quality of the reconstructed shapes.
Deep clustering aims to cluster unlabeled data by embedding them into a subspace based on deep model. The key challenge of deep clustering is to learn discriminative representations for input data with high dimensions...
详细信息
ISBN:
(纸本)9781509060146
Deep clustering aims to cluster unlabeled data by embedding them into a subspace based on deep model. The key challenge of deep clustering is to learn discriminative representations for input data with high dimensions. In this paper, we present a deep discriminative clustering network for clustering the real-world images. We use a convolutionalautoencoder stacked with a softmax layer to predict clustering assignments. To learn a discriminative representations, the proposed approach adds discriminative loss as embedded regularization with relative entropy minimization. With the discriminative loss, the network can not only produce clustering assignments, but also learn discriminative features by reducing intra-cluster distance and increasing inter-cluster distance. We evaluate the proposed method on three datasets: MNIST-full, YTF and FRGC-v2.0. We outperform state-of-the-art results on MNIST-full and FRGCv2.0 and achieve competitive result on YTF. The source code has been made publicly available at https://***/shaoxuying/DeepDiscriminativeClusteringNetwork.
In machine vision, image processing technology is the basis of target recognition and positioning. When the background of the image is complex, especially when the background feature is similar to the target feature, ...
详细信息
ISBN:
(纸本)9781538649916
In machine vision, image processing technology is the basis of target recognition and positioning. When the background of the image is complex, especially when the background feature is similar to the target feature, the accuracy of the target recognition by traditional image processing methods cannot be guaranteed. In this paper, based on the background of automatic welding technology, proposing a new method of combining the neural networks and machine vision. Specifically, the image is preprocessed by using an improved convolutional auto-encoder to enhance the target features and remove the characteristics of the main interferers. Then, use the improved traditional image processing technology to extract the target and complete the processing of the featureless image. Finally, use a binocular camera to achieve accurate positioning of the target. This paper provides a new idea for the identification and positioning of the target.
Image hashing approaches map high dimensional images to compact binary codes that preserve similarities among images. Although the image label is important information for supervised image hashing methods to generate ...
详细信息
ISBN:
(纸本)9783030033354;9783030033347
Image hashing approaches map high dimensional images to compact binary codes that preserve similarities among images. Although the image label is important information for supervised image hashing methods to generate hashing codes, the retrieval performance will be limited according to the performance of the classifier. Therefore, an effective supervised auto-encoder hashing method (SAEH) is proposed to generate low dimensional binary codes in a point-wise manner through deep convolutional neural network. The auto-encoder structure in SAEH is designed to simultaneously learn image features and generate hashing codes. Moreover, some extra relaxations for generating binary hash codes are added to the objective function. The extensive experiments on several large scale image datasets validate that the auto-encoder structure can indeed increase the performance for supervised hashing and SAEH can achieve the best image retrieval results among other prominent supervised hashing methods.
This paper presents a novel, highly-adaptable Java framework N-light-N, for the work with deep neural networks, especially with convolutional auto-encoders (CAE). While the most popular deep learning libraries focus o...
详细信息
ISBN:
(纸本)9781509009817
This paper presents a novel, highly-adaptable Java framework N-light-N, for the work with deep neural networks, especially with convolutional auto-encoders (CAE). While the most popular deep learning libraries focus on fast processing and high performance, they only implement the main-stream network architectures and network units. In recent research in the document domain, however, we have shown that modified networks, units, and training processes significantly improve the performance in various tasks. To enable the document research community with such capabilities, in this paper we introduce a novel, publicly available Deep Learning framework which is easy to use, adapt, and extend. Furthermore, we present successful applications for three tasks, including two in the domain of handwritten historical documents, and show how the framework can be used for adaptation, optimization, and deeper analysis.
We introduce the concept of diverse activation functions, and apply them into convolutional auto-encoder (CAE) to develop diverse activation CAE (DaCAE), which considerably reduces the reconstruction loss. In contrast...
详细信息
ISBN:
(纸本)9781538618295
We introduce the concept of diverse activation functions, and apply them into convolutional auto-encoder (CAE) to develop diverse activation CAE (DaCAE), which considerably reduces the reconstruction loss. In contrast to vanilla CAE only with activation functions of the same types, DaCAE incorporates diverse activations by considering their cooperation and location. In terms of the reconstruction capability, DaCAE significantly outperforms vanilla CAE and full connected auto-encoder, and we conclude rules of thumb on designing diverse activations networks. Based on the high quality of the latent bottleneck features extracted from Da-CAE, we demonstrate a satisfying advantage that fuzzy rules classifier performs better than softmax layer in supervised learning. These results could be seen as new research points in the attempts at using diverse activations to train deep neural networks and combining fuzzy inference systems with deep learning.
暂无评论