The purpose of this paper is to assess the performance of several popular features for handwriting on Mongolian offline handwriting recognition system (HWR). They have been classified into handcrafted and automaticall...
详细信息
ISBN:
(纸本)9781728121062
The purpose of this paper is to assess the performance of several popular features for handwriting on Mongolian offline handwriting recognition system (HWR). They have been classified into handcrafted and automatically learned features. The handcrafted features are distribution feature, concavity feature, local gradient histogram(LGH) feature and transforms feature. The automatically learned features are extracted by Restricted Boltzmann Machine (RBM) and autoencoder. In this paper, the handwriting recognition system is based on hybrid architectures of hidden Markov models (HMMs)-deep neural networks (DNN) which play state of art role on speech recognize (ASR) tasks. In order to performance comparison, several experiments based on different features extracted from MHW database were performed. The best system on word error rate is based on the LGH feature (5.90%), followed by the autoencoder feature (6.42%).
Recently, deep learning for physical layer has been modeled using autoencoders to model the entire communication system end-to-end. We extend these methods to improve the overall performance by adopting various learni...
详细信息
ISBN:
(纸本)9781728137155
Recently, deep learning for physical layer has been modeled using autoencoders to model the entire communication system end-to-end. We extend these methods to improve the overall performance by adopting various learning strategies when multiple users try to communicate over a shared channel. We consider cooperative and non-cooperative schemes in the 2-user Gaussian interference channel. Additionally, a simple neural network architecture is provided for wireless communication systems where channel gain matrix either attenuates or, in some cases, results in fading of the message signal sent by the transmitter.
Automated disease detection in videos and images from the gastrointestinal (GI) tract has received much attention in the last years. However, the quality of image data is often reduced due to overlays of text and posi...
详细信息
ISBN:
(纸本)9781728123424
Automated disease detection in videos and images from the gastrointestinal (GI) tract has received much attention in the last years. However, the quality of image data is often reduced due to overlays of text and positional data. In this paper, we present different methods of preprocessing such images and we describe our approach to GI disease classification for the Kvasir v2 dataset. We propose multiple approaches to inpaint problematic areas in the images to improve the anomaly classification, and we discuss the effect that such preprocessing does to the input data. In short, our experiments show that the proposed methods improve the Matthews correlation coefficient by approximately 7% in terms of better classification of GI anomalies.
Current video quality control entails a manual review of every frame for every video for pixel errors. A pixel error is a single or small group of anomalous pixels displaying incorrect colors, arising from multiple so...
详细信息
ISBN:
(纸本)9781450363174
Current video quality control entails a manual review of every frame for every video for pixel errors. A pixel error is a single or small group of anomalous pixels displaying incorrect colors, arising from multiple sources in the video production pipeline. The detection process is difficult, time consuming, and rife with human error. In this work, we present a novel approach for automated pixel error detection, applying simple machine learning techniques to great effect. We use an autoencoder architecture followed by statistical post-processing to catch all tested live action pixel anomalies while keeping the false positive rate to a minimum. We discuss previous dead pixel detection methods in image processing, and compare to other machine learning approaches.
Dimension of an inertial manifold for a chaotic attractor of spatially distributed system is estimated using autoencoder neural network. The inertial manifold is a low dimensional manifold where the chaotic attractor ...
详细信息
ISBN:
(数字)9781510628250
ISBN:
(纸本)9781510628250
Dimension of an inertial manifold for a chaotic attractor of spatially distributed system is estimated using autoencoder neural network. The inertial manifold is a low dimensional manifold where the chaotic attractor is embedded. The autoencoder maps system state vectors onto themselves letting them pass through an inner state with a reduced dimension. The training processes of the autoencoder is shown to depend dramatically on the reduced dimension: a learning curve saturates when the dimension is too small and decays if it is sufficient for a lossless information transfer. The smallest sufficient value is considered as a dimension of the inertial manifold, and the autoencoder implements a mapping onto the inertial manifold and back. The correctness of the computed dimension is confirmed by its remarkable coincidence with the one obtained as a number of covariant Lyapunov vectors with vanishing pairwise angles. These vectors are called physical modes. Unlike never having zero angles residual ones they are known to span a tangent subspace for the inertial manifold.
Content Delivery Networks (CDNs) provide high quality of service by storing content in edge-servers close to users. Attacks against CDN edge-servers can lead to loss in revenue and reputation. Attacks are becoming mor...
详细信息
ISBN:
(纸本)9783903176157
Content Delivery Networks (CDNs) provide high quality of service by storing content in edge-servers close to users. Attacks against CDN edge-servers can lead to loss in revenue and reputation. Attacks are becoming more sophisticated, and new attacks are being introduced constantly. In our previous work, we developed a security orchestration system driven by high-level security policies to dynamically deploy mitigation services. In this system, security policies are triggered at the occurrence of low-level alerts that correspond to misuse of an edge-server's resources. However, a network operator must know the effects of any attack on resources to deploy an appropriate mitigation service. Moreover, pin-pointing the actual cause (e.g., malicious IPs) of resource misuse is challenging. Also, edge-server's resources may not be affected by some attacks. Leveraging advanced machine learning techniques, we extend our system to detect new and sophisticated attacks. The goal is to enable the network operator to specify higher-level security policies without worrying about analyzing low-level resource usage alerts. Further, policy enforcement can trigger the deployment of mitigation services only for malicious entities identified by the alerts. In this perspective, we propose a Hybrid Classification Clustering (HCC) method that not only detects known sophisticated attacks accurately (with 99:9% detection recall) but is capable of detecting new attacks (with 56:4% detection recall). Further, to improve the detection rate of new attacks and anomalies, we propose an autoencoder-based Network Anomaly Detection (ANAD) method using a fully-connected autoencoder model. The evaluation results show that our model achieves 76:7% recall surpassing the isolation forest and the local outlier factor methods.
We are focusing on image classification in industrial processing taking into account the most problematic issue of the processing: the lack of labeled data. Here, we are considering three datasets: the first one is an...
详细信息
ISBN:
(纸本)9783030205188;9783030205171
We are focusing on image classification in industrial processing taking into account the most problematic issue of the processing: the lack of labeled data. Here, we are considering three datasets: the first one is an unsorted collection of all types of manufactured products and includes 100 images per class. The second one consists of products sorted into particular classes by a specialized employee and includes only ten images per class. The last one includes a massive volume of labeled images, but it is used only for the proposal validation. As the configuration is challenging for neural networks, we propose to use Image Represented by a Fuzzy Function in order to enrich original image information. We solve the task using various autoencoder architectures and prove that such the proposal increases the autoencoders success rate.
Nowadays, an intelligent transportation system (ITS) is extremely important. High accuracy traffic prediction has been studied in large-scale networks. Deep learning methods are able to analyze the complex, multi-dime...
详细信息
ISBN:
(纸本)9781728121284
Nowadays, an intelligent transportation system (ITS) is extremely important. High accuracy traffic prediction has been studied in large-scale networks. Deep learning methods are able to analyze the complex, multi-dimensional data to provide the least traffic prediction error. Both spatial and temporal dependencies provide significant implications for traffic prediction. Hence, we propose a combination of deep learning method architectures which consist of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) to analyze spatial and temporal features and predict traffic speed in multiple steps. An accident has a major impact on local traffic congestion. We employ an autoencoder with a monotonic attention mechanism to learn accident embedding input to detect unexpected accidents and their effects. An experiment was conducted on two datasets of Highways England and Thai Expressways. Results showed that spatial and temporal features can solve problems at different locations in the traffic network using the deep learning method to improve traffic prediction performance. Furthermore, a monotonic attention mechanism using an autoencoder to learn how accidents impact on traffic congestion can minimize the prediction error.
Nowadays, floating-point temporal-spatial datasets are routinely generated from scientific observational apparatuses or computer simulations at an unprecedented pace. The sheer amount of these large volumetric dataset...
详细信息
ISBN:
(纸本)9781728108582
Nowadays, floating-point temporal-spatial datasets are routinely generated from scientific observational apparatuses or computer simulations at an unprecedented pace. The sheer amount of these large volumetric datasets on the order of terabytes or petabytes consume massive resources in terms of bandwidth, storage and computational power. On the other hand, scientists, equipped with low-end post-analysis machines, often find it impossible to visualize and analyze these massive datasets with such limited resources in hand, not to mention their ultimate goal of real time analysis and visualization. To solve this discrepancy, a compact data representation has to be generated and a trade-off between resource consumption and analytical precision has to be found. There are many existing volumetric representation generating methods, almost all of which adopts some kind of hand-engineered heuristics to extract the effective portion of the datasets. However, the trade-off between resource consumption and analytical quality could not be well established due to the introduction of hand-engineered heuristics. In this paper, we present a deep learning based method that can adaptively capture the inherently complicated dynamics of temporal-spatial volumetric datasets without introducing any hand engineered features. We train an autoencoder based neural network with quantization and adaptation. Compared with existing methods, our method could learn data representation at a much lower compressed/uncompressed rate while preserving the details of original datasets. Also, our method could adapt with different data distribution and conduct compression and decompression in real time. Through extensive experiments, we show the effectiveness and efficiency of our approach over existing methods.
In this paper we address the problem of potato blemish classification and localization. A large database with multiple varieties was created containing 6 classes, i.e., healthy, damaged, greening, black dot, common sc...
详细信息
ISBN:
(纸本)9789897583513
In this paper we address the problem of potato blemish classification and localization. A large database with multiple varieties was created containing 6 classes, i.e., healthy, damaged, greening, black dot, common scab and black scurf. A Convolutional Neural Network was trained to classify face potato images and was also used as a filter to select faces where more analysis was required. Then, a combination of autoencoder and SVMs was applied on the selected images to detect damaged and greening defects in a patch-wise manner. The localization results were used to classify the potato according to the severity of the blemish. A final global evaluation of the potato was done where four face images per potato were considered to characterize the entire tuber. Experimental results show a face-wise average precision of 95% and average recall of 93%. For damaged and greening patch-wise localization, we achieve a False Positive Rate of 4.2% and 5.5% and a False Negative Rate of 14.2% and 28.1% respectively. Concerning the final potato-wise classification, we achieved in a test dataset an average precision of 92% and average recall of 91%.
暂无评论