Search

Recently, many efforts have been concentrated on land use land cover (LULC) classification due to rapid urbanization, environmental pollution, agriculture drought, frequent floods, and climate change. However, various aspects have attracted hyperspectral imaging due to there being informative discriminative features, such as spectral-spatial features. To this end, this paper is a comprehensive and systematic review of LULC classification using hyperspectral images by reviewing four significant research investigations. Moreover, the four investigations have addressed the following points: (1) the main components of the hyperspectral imaging, the modes of hyperspectral imaging with data acquisition methods, and the intrinsic differences between hyperspectral image and multispectral image, (2) the role of machine learning in LULC classification, and the standard deep learning methods: Convolution Neural Network (CNN), Stacked Autoencoder (SAE), Deep Belief Network (DBN), Recurrent Neural Network (RNN), and Generative Adversarial Network (GAN), (3) the standard benchmark hyperspectral datasets and the evaluation criteria, (4) the main challenges of LULC classification with the possible solutions for the limited training samples issue, the promising future directions, and finally the recent applications for LULC classification. [ABSTRACT FROM AUTHOR]

[Display omitted] • We propose a novel distillation approach, i.e., LGAD, which uses structure hints in lane labels to guide the attention of lane segmentation networks to better capture long-range textural information. • We carefully investigate the inner mechanism of LGAD, the considerations of choosing among different layer mimicking paths, and the optimization strategies. • Extensive experiments on two benchmark datasets demonstrate the effectiveness of the proposed LGAD on boosting the performance of lane segmentation networks. Contemporary segmentation methods are usually based on deep fully convolutional networks (FCNs). However, the layer-by-layer convolutions with a growing receptive field is not good at capturing long-range contexts such as lane markers in the scene. In this paper, we address this issue by designing a distillation method that exploits label structure when training segmentation network. The intuition is that the ground-truth lane annotations themselves exhibit internal structure. We broadcast the structure hints throughout a teacher network, i.e., we train a teacher network that consumes a lane label map as input and attempts to replicate it as output. Then, the attention maps of the teacher network are adopted as supervisors of the student segmentation network. The teacher network, with label structure information embedded, knows distinctly where the convolutional layers should pay visual attention into. The proposed method is named as Label-guided Attention Distillation (LGAD). It turns out that the student network learns significantly better with LGAD than when learning alone. As the teacher network is deprecated after training, our method does not increase the inference time. Note that LGAD can be easily incorporated in any lane segmentation network. To validate the effectiveness of the proposed LGAD method, extensive experiments have been conducted on two popular lane detection benchmarks: TuSimple and CULane. The results show consistent improvement across a variety of convolutional neural network architectures. Specifically, we demonstrate the accuracy boost of LGAD on the lightweight model ENet. It turns out that the ENet-LGAD surpasses existing lane segmentation algorithms. The main contributions of this paper include a newly proposed distillation training strategy (LGAD) and solid experimental investigation of the inner mechanism of LGAD. [ABSTRACT FROM AUTHOR]

Low-light image enhancement studies how to improve the quality of images captured under poor lighting conditions, which is of real-world importance. Currently, convolutional neural network (CNN)-based methods with state-of-the-art performance have become the mainstream of research. However, most CNN-based methods improve the performance of the algorithm by increasing the width and depth of the neural network, which requires large computing device resources. In this paper, we propose a knowledge distillation method for low light image enhancement. The proposed method uses a teacher-student framework in which the teacher network tries to transfer the rich knowledge to the student network. The student network learns the knowledge of image enhancement under the supervision of ground truth images and under the guidance of the teacher network simultaneously. Knowledge transfer between the teacher-student network is accomplished by distillation loss based on attention maps. We designed a gradient-guided low-light image enhancement network that can be divided into an enhancement branch and a gradient branch, where the enhancement branch is learned under the guidance of the gradient branch to better preserve structural information. The teacher and student networks use a similar structure, but they have different model sizes. The teacher network has more parameters and more powerful learning capabilities than the student network. With the help of knowledge distillation, our approach can improve the performance of the student network without increasing the computational burden during the testing phase. The qualitative and quantitative experimental results demonstrate the superiority of our method compared to the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

The success of convolutional neural networks (CNNs) has made low-latency inference services on Graphic Processing Units (GPUs) a hot research topic. However, GPUs are hardware processors with high power consumption. To have the least energy consumption while meeting latency Service-Level-Objective (SLO), batching strategy and dynamic voltage frequency scaling (DVFS) are two important solutions. However, existing studies do not coordinate them and regard CNN as a black box, which makes inference services less energy-efficient. In this paper, we propose EALI, an energy-aware layer-level adaptive scheduling framework that is comprised of a power prediction model, a layer combination strategy, and an energy-aware layer-level scheduler. The power prediction model uses classic machine learning techniques to predict fine-grained layer-level power consumption. The layer combination strategy combines multiple layers into optimization units to lower scheduling overhead and complexity. The energy-aware layer-level scheduler adaptively coordinates batching strategy and layer-level DVFS according to workloads to minimize the energy consumption while meeting SLO. Our experimental results on NVIDIA Tesla M40 and V100 GPUs show that, compared to the state-of-the-art approaches, EALI decreases energy consumption by up to 36.24% while meeting SLO. [ABSTRACT FROM AUTHOR]

• We propose GATE-CNN architecture for binary classification of brain tumor using MR (Magnetic Resonance) images. • We have implemented our model on three different datasets; all consists of MRI with a significant difference in number of images. • The first dataset consists of tumorous and non-tumorous MRIs, second consists of MRIs with glioma and pitutiary (types of brain tumor), and the third consists of cancerous and non-cancerous (severity of brain tumor) images. • For all three datasets we found better classification results than any of the existing state-of-the-art models. Early and accurate detection is a solitary precaution to overcome the brain tumor. Otherwise, it will result in a deadly disease. Brain tumor (BT) detection using magnetic resonance is an essential and challenging job in the medical domain. To combat the aggressive spreading of BT, clinical imaging (MRI and X-Ray) can be an appropriate method for diagnosis. In this paper, we applied attention based image classification, which is a new paradigm for obtaining improved accuracy than the state-of-the-art models. We propose a GATE-CNN (Graph Attention AutoEncoder-Convolution Neural Network) model in this work for the classification of BT. We calculate the attention values of the neighboring pixels on each and every pixel present in the graph then process the graph using GATE framework and the processed graph with attention values is then passed to CNN framework for generation of final output. Further, we apply Adamax optimizer to optimize the training hyperparameters of CNN. We perform the BT classification using the proposed method on three different datasets among benign and malignant BT for first dataset, glioma and pituitary for second dataset, and normal and abnormal BT images for third dataset; all the three datasets consists of MRI images. We then compare the proposed model with a variety of CNN (Deep-CNN, Multi-Input CNN, and basic CNN) models dealing with several types of medical imaging (thermal image, MRI, X-Ray, etc.). Our model results in the accuracy of 98.27% for first dataset, 99.83% for second dataset, and 98.78% for third dataset demonstrating preferable network performance than the already established state-of-the-art CNN models. [ABSTRACT FROM AUTHOR]

Evolutionary Neural Architecture Search (ENAS) is a promising method for the automated design of deep network architecture, which has attracted extensive attention in the field of automated machine learning. However, the existing ENAS methods often need a lot of computing resources to design CNN architecture automatically. In order to achieve efficient and automated design of CNNs, this paper focuses on two aspects to improve efficiency. On the one hand, efficient CNN-based building blocks are introduced to ensure the effectiveness of the generated architectures and a triplet attention mechanism is incorporated into the architectures to further improve the classification performance. On the other hand, a random forest-based performance predictor is used in the fitness evaluation to reduce the amount of computation required to train each individual from scratch. Experimental results show that the proposed algorithm can significantly reduce the computational resources required and achieve competitive classification performance on the CIFAR dataset. Also, the architecture designed for the traffic sign recognition task exceeds the accuracy of manual expert design. [ABSTRACT FROM AUTHOR]

As a powerful machine learning method, deep learning has attracted the attention of numerous researchers. While exploring a high-performance neural network model, the floating-point operations of a neural network model are also increasing. In recent years, many researchers have noticed that efficiency is also one of important indicators to measure the property of neural network models. Obviously, the efficient neural network model is more helpful to deploy on mobile and embedded devices. Therefore, the efficient neural network model becomes a hot research spot. In this paper, we review the methods related to the structural design of efficient convolution neural networks in recent years. According to the characteristics of these methods, we divide them into three kinds of methods: model pruning, efficient architecture, and neural architecture search. Detailed analyses of each method are presented to demonstrate their advantages and disadvantages. Then, we comprehensively compare them in detail and propose many suggestions about the design of the efficient convolution neural network model structure. Inspired by these suggestions, we built a new efficient neural network model, SharedNet. And the SharedNet obtains the best accuracy of manually-designed efficient CNN models on the ImageNet dataset. [ABSTRACT FROM AUTHOR]

This paper shows a computational model that classifies Cybergrooming attacks in the context of COP (child online protection) using Natural Language Processing (NLP) and Convolutional Neural Networks (CNN). The model predicts a high number of false positives, therefore low precision and F-score, but a high accuracy. In this issue, where the number of messages in the context of grooming are so low compared to the number of conversations and messages from other contexts, it can be concluded that is a very consistent and useful result as it captures a high number of true positives, considering that the classifier works for messages. Performing the training of machine learning algorithms with neural networks, semantic analysis and NLP, allows approximate representation of knowledge contributing to discovery of pseudo-intelligent information in these environments and reducing human intervention for characterization of underlying abnormal behavior and detecting messages that potentially represent these attacks. [ABSTRACT FROM AUTHOR]

• A CNN for heavy rain removal is proposed based on two networks connected in series. • Novel attention modules are introduced to effectively remove heavy rain effect. • New heavy rain data is proposed with new haze map and various heavy rain models. In this paper, we propose a novel heavy rain removal algorithm using a deep neural network. Unlike most of the existing deraining methods, heavy rain removal is a more challenging task because it is necessary to remove both the rain marks and the haze effects, which are entangled in a complex manner. Motivated by this, we propose a new end-to-end two-stage attention network for single-shot heavy rain removal. The proposed network is connected serially with a removal network and a recovery network, which are based on a newly introduced dilation-wise attention block and skip attention block. Based on these attention techniques, the removal network predicts the heavy rain effect that needs to be removed from a given image, and the recovery network successfully predicts the details that need to be recovered, resulting in a clean image. We also introduce a new realistic RainCityscapes+ dataset, composed of synthesized outdoor images, and demonstrate extensive experiments, the results of which show our approach outperforms the state-of-the-art methods on both real and synthetic datasets quantitatively and qualitatively. [ABSTRACT FROM AUTHOR]

Neuroevolution is a machine learning method for evolving neural networks parameters and topology with a high degree of flexibility that makes them applicable to a wide range of architectures. Neuroevolution has been popular in reinforcement learning and has also shown to be promising for deep learning. The major feature of Bayesian optimisation is in reducing computational load by approximating the actual model with an acquisition function (surrogate model) that is computationally cheaper. A major limitation of neuroevolution is the high computational time required for convergence since learning (evolution) typically does not utilize gradient information. Bayesian optimisation, which is also known as surrogate-assisted optimisation, has been popular for expensive engineering optimisation problems and hyper-parameter tuning in machine learning. It has potential for training deep learning models via neuroevolution given large datasets and complex models. Recent advances in parallel and distributed computing have enabled efficient implementation of neuroevolution for complex and computationally expensive neural models. In this paper, we present a Bayesian optimisation framework for deep neuroevolution using a distributed architecture to provide computational efficiency in training. Our results demonstrate promising results for simple to deep neural network models such as convolutional neural networks which motivates further applications. [ABSTRACT FROM AUTHOR]

• Trait agnostic generic approach for constructing deep cancelable biometric templates. • Deep incremental cancelable system. • Similarity attack resistant biometric templates. • Phase-wise incremental learning enabled cancelable system. In a post-COVID-19 world, extensive study of deep learning-based biometric authentication techniques prompts the need to secure them. Further, the biometric data is assumed to be largely immutable; thus, if it is compromised, it is lost forever. Hence, reliable and secure biometric authentication is of utmost importance. In this paper, we address the security and privacy concerns of biometric templates generated via deep networks. We propose a cancelable biometric authentication approach. The framework consists of a lightweight Convolutional Neural Network (CNN) with a few-shot enrollment for generating biometric templates. Further, to enhance biometric templates' discriminative power and to provide revocability, biometric templates are projected onto a random subspace (based on the user-specific key). Later projected biometric templates are mapped onto robust n - bit unique codes (using a KNN classifier) and protected via. SHA-3 hash digest. Moreover, a real-world biometric authentication system is always dynamic (users keep on changing). Thus we have also integrated phase-wise incremental learning within a deep learning-based cancelable biometric authentication framework. This is the first work in which deep cancelable templates are generated incrementally to the best of our knowledge. We analyze the proposed scheme for its performance and privacy preservation on three benchmarks constrained iris data-sets and over one unconstrained iris data-set along with one publicly available knuckle data-set. Furthermore, it has been demonstrated that the proposed cancelable incremental framework strictly follows the four fundamental properties of cancelability viz. non-invertibility, unlinkability, revocability, and usability. [ABSTRACT FROM AUTHOR]

• A time–frequency deep metric learning model is proposed for MTS classification. • A consistency regularizer is designed to capture the correlations among levels. • An effective optimization strategy is developed to train the model hierarchically. • Results on 18 MTS datasets show that our method outperforms other methods. Multivariate time series (MTS) data exist in various fields of studies and MTS classification is an important research topic in the machine learning community. Researchers have proposed many MTS classification models over the years and the distance-based methods along with nearest neighbor classifier achieve good performance. However, the current methods mainly focus on defining distance metric on time-domain of MTS and ignore frequency information. Besides, these methods usually define the same linear distance metric for different datasets, which is not suitable for capturing the nonlinear relationship of MTS and degrades the discriminative power of the distance metric. In this paper, we propose a time–frequency deep metric learning (TFDM) approach for MTS classification. The multilevel discrete wavelet decomposition is first adopted to decompose an MTS into a group of sub-MTS so as to extract multilevel time–frequency representations. Then, a deep convolutional neural network is developed for each level to learn level-specific nonlinear features and a metric learning layer is added on the top of the network to learn the semantic similarity of MTS. Moreover, a cross-level consistency regularization term is designed to encourage the distance metrics of different levels to be consistent for capturing the correlations among different levels. Finally, we use 1-nearest neighbor to classify MTS according to the learned distance metrics. Extensive experiments on 18 benchmark datasets show the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

Database security focuses on protecting most organization's virtual data storage unit and confidential information from malicious threats and external attacks. To keep out data secure, we need to use a role-based access control (RBAC) approach to accurately differentiate access permissions, but SQL queries written by an authorized user have very similar characteristics and are difficult to distinguish. In this paper, we propose a method of optimizing CNN-LSTM neural networks with particle swarm optimization (PSO) to classify the roles in RBAC system. Convolutional neural network (CNN) can extract parsed SQL queries into smaller details and features through an analysis mechanism. Long short-term memory (LSTM) is also suitable for modeling the temporal information of SQL queries to recognize the context of user authorities. PSO repeatedly searches and optimizes the complex hyperparameter space of the CNN-LSTM. Our PSO-based CNN-LSTM neural networks outperform other deep learning and machine learning models in the TPC-E benchmark SQL query statement. Finally, experiments and analysis show the usefulness of PSO and identify the important SQL query features that affect user role classification. [ABSTRACT FROM AUTHOR]

A novel sparse tensor dictionary learning algorithm and a convolutional neural network (CNN) classification method based on this algorithm are proposed for hyperspectral image (HSI) in this paper. The HSI is intuitively represented as a three-dimensional (3-D) cube, and utilizing the joint spatial-spectral information can significantly improve the accuracy of HSI classification. Therefore, the proposed atom-substituted tensor dictionary learning (ASTDL) algorithm utilizes tensor techniques to extract 3-D joint spatial-spectral features from HSI cubes directly. Sparsity constraint is enforced on the coefficient tensors, which obeys the sparsity attribute of HSI. The proposed ASTDL enhanced CNN (ASTDL-CNN) classification method utilizes a two-dimensional (2-D) CNN to extract deep features from the feature tensors obtained by the ASTDL, and to perform the pixel-wise classification. The use of ASTDL, which extracts intrinsic tensor features before CNN, alleviates within-class spatial-spectral variation and reduces the requirement of CNN for the labeled data. Whereas, the CNN works as a 3-D classifier, and provides the final classification results. Besides, we perform the majority vote on the classification map obtained by CNN to refine the classification. The performance of the proposed method is evaluated on three real HSI data sets. The competitive results to the compared state-of-the-art methods demonstrate that the ASTDL-CNN can provide accurate and robust classification results. [ABSTRACT FROM AUTHOR]

Extreme Learning Machine (ELM) has gained lots of research interests due to its universal approximation capability and fast learning speed. Although several prior works have focused on developing deep ELM, it is still an open problem to design effective deep ELM. Stacking random layers will result in overfitting and accumulation of random errors. To address this issue, this paper presents a simple yet effective deep ELM called Densely Connected Convolutional ELM (DC2ELM) for hyperspectral image spectral-spatial classification. First, we introduce dense connection into ELM to make full use of intermediate feature maps produced by randomized convolutional layers, which is beneficial to reduce the random error. Secondly, stacked ELM auto-encoders are employed to generate reduced representation, leading to a deeper architecture. The proposed approach consists of fewer trainable parameters than traditional convolutional neural networks and can easily be trained without any iterative parameters tuning, making it easier to implement and apply in practice. We compare the proposed approach with many prior arts over three real hyperspectral images, showing that the proposed approach can achieve superior performance using limited training data and with a reduced risk of overfitting the training data. [ABSTRACT FROM AUTHOR]

Recent years has witnessed the success of convolutional neural networks (CNNs) in many machine learning and pattern recognition applications, especially in image recognition. However, due to the increasing model complexity, the parameter redundancy problem arises, and greatly degrades the performance of CNNs. To alleviate this problem, various regularization techniques, such as Dropout, have been proposed and proved their effectiveness. In this paper, we propose a novel adaptive kernel-based weight decorrelation (AKWD) framework, in order to regularize CNNs for better generalization. Different from existing works, the correlation between paring weights is measured by the cosine distance defined in RKHS associated with a specific kernel. The case with the well-known Gaussian kernel is investigated in detail, where the bandwidth parameter is adaptively estimated. By regularizing CNN models of different capacities using AKWD, better performance is achieved on several benchmark databases for both object classification and face verification tasks. In particular, when Dropout or BatchNorm is present, even higher improvements are obtained using the proposed AKWD, that demonstrates a good compatibility of the proposed regularizer with other regularization techniques. [ABSTRACT FROM AUTHOR]

• MSDM is proposed to obtain primitive visual context for non-rigid salient objects. • CWAM is designed for channels boosting and suppression. • The framework achieves the state-of-the-art performance. Contextual information has played an important role in salient object detection. However, due to the fixed geometric structures of convolution kernels employed by existing Convolutional Neural Networks (CNNs) based methods, it is difficult to extract meaningfully visual contexts for those salient objects with varying sizes and non-rigid shapes. To address this problem, in this paper, we propose a Multi-Scale Deformation Module (MSDM) to capture multi-scale visual cues and varying shapes of salient objects. Moreover, most existing CNNs based methods treat all channels of feature maps equally, which tends to differ from the fact that different channels actually contribute differently to saliency prediction. For that, we involve a novel Channel-Wise Attention Mechanism (CWAM) after MSDM to highlight those informative channels while suppressing those confusing ones. Experimental results on five benchmark datasets demonstrate the superiority of the proposed method over the state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) has incorporated various automatic systems and frameworks to diagnose the severity of depression using hand-crafted features. However, process of feature selection needs domain knowledge and is still time-consuming and subjective. Deep learning technology has been successfully adopted for depression recognition. Most previous works pre-train the deep models on large databases followed by fine-tuning with depression databases (i.e., AVEC2013, AVEC2014). In the present paper we propose an integrated framework – Deep Local Global Attention Convolutional Neural Network (DLGA-CNN) for depression recognition, which adopts CNN with attention mechanism as well as weighted spatial pyramid pooling (WSPP) to learn a deep and global representation. Two branches are introduced: Local Attention based CNN (LA-CNN) focuses on the local patches, while Global Attention based CNN (GA-CNN) learns the global patterns from the entire facial region. To capture the complementary information between the two branches, Local–Global Attention-based CNN (LGA-CNN) is proposed. After feature aggregation, WSPP is used to learn the depression patterns. Comprehensive experiments on AVEC2013 and AVEC2014 depression databases have demonstrated that the proposed method is capable of mining the underlying depression patterns of facial videos and outperforms the most of the state-of-the-art video-based depression recognition approaches. [ABSTRACT FROM AUTHOR]

Optomotor response (OMR) describes an innate orienting behavior for numerous kinds of model animals when the surrounding visual scene is moving, and is often used for evaluating animals' visual function and nervous system. OMR measurements are generally performed and analyzed by skillful operators, but this manual measurement suffers from low accuracy and efficiency, which further yields unreliable and even misleading results. In this paper, we present a quantitative method to automatically measure the OMR in mice. The presented measurement method consists of a set of algorithms, e.g., convolutional neural network (CNN) to track the orientation of the mouse's head, and a feature extraction method to measure the OMRs. Compared with existing techniques, the proposed method can measure the exact starting and ending time instants of each OMR in a video segment of a mouse motion record, based on the visual detection and analysis algorithms for mice motion, which is more precise and consistent compared to only judging the presence or absence of OMR for a whole segment of a motion record in almost all the existing OMR measuring methods. In the experiments, 21 mice (7 wt mice, 7 OPTNE50K mice, and 7 OPTNE50K + BMCs mice) with three levels of optic nerve injury treatments are performed the OMR tests. The observed statistical difference in results of the three groups of mice with different vision verifies the validity of the method. Compared with the calibration results from ophthalmological experts, the system can achieve the recognition rate of 94.89%. The proposed method also provides a quantitative and analytical alternative of the OMR behavior of mice in neuroscience and ophthalmology studies. [ABSTRACT FROM AUTHOR]

• An efficient algorithm for ECG heartbeats classification is proposed. The proposed algorithm termed ML-WCNN is based on Convolutional Neural Networks from the deep learning theory, and the Stationary Wavelet Transform (SWT). • The proposed End-to-End deep learning architectures can automatically learn effective discriminative features from the original ECG signal and its wavelets coefficients simultaneously. • The developed ML-WCNN methods can recognize effectively six ECG beat classes without requirement of any pre-processing or features extraction steps. • ECG signals form the standard database MIT-BIH Arrhythmia have been used for evaluating the proposed approach in comparison with others established Machine and Deep learning techniques. • The quantitative results show that the proposed method exhibits a higher or comparable performances for ECG classification on the against some state of the art algorithms with a maximum accuracy of 99.57%. This paper presents a new End-to-End Deep Learning method for heart diseases diagnosis from single channel ECG signal. Motivated by the great efficiency and popularity of deep learning algorithms for time series classification, the proposed work is mainly based on the One Dimensional Convolutional Neural Networks (1D-CNN). Unlike the traditional CNN models based classification, a new Multi-Level Wavelet Convolutional Neural Networks (ML-WCNN) is proposed to recognize automatically various types of cardiac arrhythmias. The proposed approach incorporates the 1D-CNN model and the Stationary Wavelet Transform (SWT) to extract discriminative features from different wavelet sub-bands and from the raw ECG signal simultaneously. The extracted features by the ML-WCNN model are then merged using different fusion strategies, especially by concatenation and maximization. This improves greatly the features learning process at different scales of the ECG signal, providing better diagnosis performances. The proposed ML-WCNN framework is evaluated on the Standard Database MIT-BIH Arrhythmia considering six classes of heart Beats: Normal (N), Premature Ventricular Contraction (PVC), Right Bundle Brunch Block (RBBB), Left Bundle Brunch Block (LBBB), Atrial Premature Contraction (APC) and Paced beats (PAC). The experimental results demonstrate the superiority of the proposed ML-WCNN, in comparison with the state-of-the-art heart diseases diagnosis based Machine/Deep learning algorithms, with a maximum Accuracy of 99.57 % using the 10-fold cross validation technique. [ABSTRACT FROM AUTHOR]

This paper presents a hybrid fuzzy integrated convolutional neural network (HFICNN) to deal with the similarity feature recognition problem in abnormal netflow detection, which is that different labels with the same netflow eigenvalues are not well disposed by traditional machine learning and deep learning methods. The HFICNN model can convert the same feature individuals with different labels to different feature individuals with different labels by constructing an integration method. Our method recognizes the actual environment and operational complexity, and converts the unique occurrence time feature performance as a new feature for feature recognition. HFICNN integrates time-dependent records into a new record to identify a feature, and converts the integrated records to the original label. The integration process plays a good role and effect in the entire process. HFICNN is realized with the aid of function mapping label redefinition (FMLR), adaptive feature integration convolutional neurons (AFICNs), fuzzy classification (FC), and inverse function mapping integrated reduction (IFMIR). The ICMPv6-based DDoS attacks dataset of a new-generation network is tested, and experimental results show that HFICNN performs better than 10 types of traditional machine learning and two types of deep learning methods on the similarity feature recognition problem, and the HFICNN model is reliable and effective. [ABSTRACT FROM AUTHOR]

Deep learning plays an important role in the field of machine learning. However, deterministic methods such as neural networks cannot capture the model uncertainty. Bayesian neural network (BNN) are recently under consideration since Bayesian models provide a theoretical framework to infer model uncertainty. Since it is often difficult to find an analytical solution for BNNs, an effective and efficient approximate inference method is very important for model training and prediction. The generalized version of expectation propagation (GEP) was recently proposed and considered a powerful approximate inference method, which is based on the minimization of Kullback–Leibler (KL) divergence of the true posterior and the approximate distributions. In this paper, we further instantiate the GEP to provide an effective and efficient approximate inference method for BNNs. We assess this method on BNNs including fully connected neural networks and convolutional neural networks on multiple benchmark datasets and show a better performance than some state-of-the-art approximate inference methods. [ABSTRACT FROM AUTHOR]

Early prediction of Alzheimer's disease (AD) is crucial for delaying its progression. As a chronic disease, ignoring the temporal dimension of AD data affects the performance of a progression detection and medically unacceptable. Besides, AD patients are represented by heterogeneous, yet complementary, multimodalities. Multitask modeling improves progression-detection performance, robustness, and stability. However, multimodal multitask modeling has not been evaluated using time series and deep learning paradigm, especially for AD progression detection. In this paper, we propose a robust ensemble deep learning model based on a stacked convolutional neural network (CNN) and a bidirectional long short-term memory (BiLSTM) network. This multimodal multitask model jointly predicts multiple variables based on the fusion of five types of multimodal time series data plus a set of background (BG) knowledge. Predicted variables include AD multiclass progression task, and four critical cognitive scores regression tasks. The proposed model extracts local and longitudinal features of each modality using a stacked CNN and BiLSTM network. Concurrently, local features are extracted from the BG data using a feed-forward neural network. Resultant features are fused to a deep network to detect common patterns which jointly used to predict the classification and regression tasks. To validate our model, we performed six experiments on five modalities from Alzheimer's Disease Neuroimaging Initiative (ADNI) of 1536 subjects. The results of the proposed approach achieve state-of-the-art performance for both multiclass progression and regression tasks. Moreover, our approach can be generalized in other medial domains to analyze heterogeneous temporal data for predicting patient's future status. [ABSTRACT FROM AUTHOR]

This paper aims to address the human action recognition issue by using convolutional long short-term memory networks (Conv-LSTM) and fully-connected LSTM (FC-LSTM) with different attentions. To this end, the spatial-temporal dual-attention network (STDAN), which is mainly composed of feature extraction, attention and fusion modules, is designed. Different from the features of high-level fully-connected layer mostly used in previous work, the features of convolution and fully-connected layers of convolutional neural network (CNN) are both extracted in STDAN, which can enrich the initial level of video representation. Besides, the Conv-LSTM and FC-LSTM are employed to handle the long-duration sequential features with different temporal context information. To reinforce the spatial-temporal attention ability, a temporal attention module (TAM) and a joint spatial-temporal attention module (JSTAM) are implemented. Through the principle components analysis (PCA) and features fusion, the potential of STDAN is effectively explored and weighted. Finally, the experimental results show that the proposed STDAN has better recognition performance than existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Impurity detection involves detecting small impurities in the liquid inside an opaque glass bottle with complex textures by looking through the bottleneck. Sometimes experts have to observe continuous frames to determine the existence of an impurity. In recent years, region-based convolutional neural networks have gained incremental successes in common object detection tasks. However, sequential impurity detections present more challenging issues than detecting targets in a single frame, because consecutive motions and appearance changes of impurities cannot be captured using those common object detectors. In this paper, we propose a simple and controllable ensemble architecture to alleviate this problem. Specifically, a siamese fusion network is used to generate impurity proposals, then an attention model based on visual features and trajectories is proposed to localize a unique region proposal in each frame, finally, a sequential region proposal classifier using a long-term recurrent convolutional network is applied to refine impurity detection performances. The proposed method achieves 79.81 % mAP on IML-DET datasets, outperforming a comparable state-of-the-art Mask R-CNN model. [ABSTRACT FROM AUTHOR]

Hyperparameter tuning is critical for the performance of machine learning algorithms. However, a noticeable limitation is the high computational cost of algorithm evaluation for complex models or for large datasets, which makes the tuning process highly inefficient. In this paper, we propose a novel model-based method for efficient hyperparameter optimization. Firstly, we frame this optimization process as a reinforcement learning problem and then employ an agent to tune hyperparameters sequentially. In addition, a model that learns how to evaluate an algorithm is used to speed up the training. However, model inaccuracy is further exacerbated by long-term use, resulting in collapse performance. We propose a novel method for controlling the model use by measuring the impact of the model on the policy and limiting it to a proper range. Thus, the horizon of the model use can be dynamically adjusted. We apply the proposed method to tune the hyperparameters of the extreme gradient boosting and convolutional neural networks on 101 tasks. The experimental results verify that the proposed method achieves the highest accuracy on 86.1% of the tasks, compared with other state-of-the-art methods and the average ranking of runtime is significant lower than all methods by using the predictive model. [ABSTRACT FROM AUTHOR]

In recent years, convolutional neural networks (CNNs) have achieved success in the field of computer vision. However, their large storage requirements and high computational consumption limit their scope of application, especially, in mobile devices. Network pruning has been proven to be an effective approach for relieving such burdens. However, the pruning ratio and accuracy of the existing methods need to be improved. In this paper, a scheme of network channel pruning, based on sparse learning and the genetic algorithm, is proposed to achieve a better balance between the pruning ratio and accuracy. A gradient-limited L 1/2 regularization is introduced into the model training, to achieve channel-level sparsity. In addition, a dynamically adjustable evaluation factor is proposed to evaluate the importance of channels and the genetic algorithm is utilized to acquire a suitable subnetwork from the sparse model. Our method is evaluated against several state-of-the-art CNNs on three different classification datasets, and is found to achieve, on an average, 63.8%/50.8%/58.6% reduction in parameters and 49.1%/42.9%/58.7% reduction in computational complexity, with negligible loss of accuracy. Our method also achieves 42.84% reduction in parameters and 30.06% reduction in computational complexity of the detection network. [ABSTRACT FROM AUTHOR]

Modeling interpretable artificial intelligence (AI) for flood forecasting represents a serious challenge: both accuracy and interpretability are indispensable. Because of the uncertainty and nonlinearity of flood, existing hydrological solutions always achieve low prediction robustness while machine learning (ML) approaches neglect the physical interpretability of models. In this paper, we focus on the need for flood forecasting and propose an interpretable Spatio-Temporal Attention Long Short Term Memory model (STA-LSTM) based on LSTM and attention mechanism. We use dynamic attention mechanism and LSTM to build model, Max-Min method to normalize data, variable control method to select hyperparameters, and Adam algorithm to train the model. Emphasis is placed on the visualization and interpretation of attention weights. Experiment results on three small and medium basins in China suggest that the proposed STA-LSTM model outperforms Historical Average (HA), Fully Connected Network (FCN), Convolutional Neural Networks (CNN), Graph Convolutional Networks (GCN), original LSTM (LSTM), spatial attention LSTM (SA-LSTM), and temporal attention LSTM (TA-LSTM) in most cases. Visualization and interpretation of spatial and temporal attention weights reflect the reasonability of the proposed attention-based model. [ABSTRACT FROM AUTHOR]

With the rapid developments of deep neural networks related applications, approaches for accelerating computationally intensive convolutional neural networks, such as network quantization, pruning, knowledge distillation, have attracted ever-increasing attention. Network binarization is an extreme form of network quantization technique, which binarizes the network weights and/or activation values to save computational resources. However, it often introduces noises into the network, and requires larger model size (more parameters) to compensate for the loss of representation capacity. To address the model complexity reduction challenges and further improve the network performance, this paper proposes an approach: slimming binarized neural networks (SBNN), which reduces complexity of binarized networks with acceptable accuracy loss. SBNN prunes the convolutional layers and fully-connected layer in a binarized network. Then it is refined by the proposed SoftSign function, knowledge distillation and full-precision computation to enhance the network accuracy. The proposed SBNN can be also conveniently applied to a pre-trained binarized network. We demonstrate the effectiveness of our approach through several state-of-the-art binarized models. For AlexNet and ResNet-18 on ILSVRC-2012 dataset, SBNN obtains negligible accuracy loss but even a better accuracy than the pre-pruning model while using only 75% of original filters. [ABSTRACT FROM AUTHOR]

The electromyogram (EMG) signals from an individual's muscles can reflect the biomechanics of human movement. The accurate classification of individual and combined finger movements using surface EMG signals is able to support many applications such as dexterous prosthetic hand control. The existing research of EMG-based hand gesture classification faces the challenges of inaccurate classification, insufficient generalization ability and weak robustness. To address these problems, this paper proposes a deep learning model that combines convolutional auto-encoder and convolutional neural network (CAE+CNN) to classify an EMG dataset consisting of 10 classes of hand gestures. The proposed method shrinks the inputs into a smaller latent space representation using CAE and the resultant compressed features are served as inputs of CNN, which reduces the redundancy of EMG signals and improves the classification accuracy and training efficiency. Besides, to enhance the robustness and generalization ability for classification, a data processing approach is proposed which combines the windowing method and majority voting of the obtained results from the classifier. In addition, comprehensive comparative study is carried out with 8 widely applied and state-of-the-art classifiers in terms of classification accuracy, robustness subject to noise and statistical analysis (sensitivity, specificity, precision, F1 Score and Matthews correlation coefficient). The results demonstrates that the integration of windowing method, CAE+CNN and majority voting achieves the best performance (99.38% test accuracy for the data without adding noise, which is 3.78% higher than the best classifier used for comparison), strongest robustness (achieved 98.13% test accuracy when Gaussian noise of level 1e-5 is added to the raw dataset, which is 4.07% higher than the best classifier used for comparison) and statistical properties compared to other classifiers, which shows the potential for healthcare applications such as movement intention detection and dexterous prostheses control. [ABSTRACT FROM AUTHOR]

The identification of chromatin accessibility is a significant part of the genomics and genetics. However, high-throughput experimental techniques are costly and impractical for systematic identification of accessibility. Many computational methods were proposed to predict the functional regions of chromatin purely relying on DNA sequences, but they could not take full advantage of sequence information to capture hidden complex motifs among DNA sequences. Recently, deep learning algorithms have been incorporated into the chromatin accessibility predication and achieved the remarkable results. Nevertheless, there still exists a problem in chromatin accessibility prediction as how to effectively represent the complex features merely from DNA sequences. Thus, developing efficient computational methods is becoming increasingly urgent to identify functional regions of the genome. In this paper, combining convolutional and gated recurrent unit neural networks with attention mechanism, we develop a discriminative computational framework DeepANF to adaptively extract hidden pattern features and identify the chromatin accessibility based on distributed representation of DNA sequences. To verify the efficacy of the DeepANF framework, we conduct extensive experiments on five large scale datasets, and experimental results reveal that our framework not only consistently outperforms these published methods for chromatin accessibility prediction tasks, but also extracts more discriminative features from pure DNA sequences than published methods, especially on MCF-7 dataset. [ABSTRACT FROM AUTHOR]

Automatic plant recognition using AI is a challenging problem. In addition to the recognition of the plant specimen, we also want to recognize the plant type in its actual living environment, which is more difficult because of the background noise. In this paper, we propose a novel method, referred to as the two-way attention model using the deep convolutional neural network. As the name implies, it has two ways of attentions: the first attention way is the family first attention , which is based on the standard plant taxonomy and aims to recognize the plant's family. Specifically, we create plant family labels as another objective of the learning under the multi-task learning framework. To deal with conflicting prediction of family and species labels, we propose an implicit tree model with a dedicated loss function to maintain the correspondence between family and species labels. The second attention way is the max-sum attention , which focuses on the discriminative features of the input image by finding the max-sum part of the fully convolutional network heat map. Because these two ways of attention are compatible, we combine both discriminative feature learning and part based attention. The experiments on four challenging datasets (i.e. Malayakew, ICL, Flowers 102 and CFH plant) confirm the effectiveness of our method – the recognition accuracy over those four dataset reaches 99.8%, 99.9%, 97.2% and 79.5% respectively. [ABSTRACT FROM AUTHOR]

Abstract Over the last decade, substantial advances have been made in various computer vision technologies and many of them are based on convolutional neural network (CNN) architecture. Typically, CNN is trained by a stochastic gradient descent algorithm using Back-Propagation (BP) but the training process is adversely affected by slow convergence and the need for extensive parameter tuning. In this paper, we propose a new CNN architecture and training algorithm based on an Extreme Learning Machine (ELM) to overcome these drawbacks. The proposed training algorithm is a layer-wise training method for CNN, and uses an alternating strategy of random convolutional filters and semi-supervised filters to combine the advantages of both approaches. On each semi-supervised layer, the CNN efficiently solves a convex optimization problem based on nonlinear random projection. It is faster and requires less human effort than BP-based training. We experimentally validated the proposed method using a well-known character and object recognition benchmark. In our experiment, the performance of our method is comparable to approaches based on deep features and has higher accuracy than other unsupervised feature-learning methods. [ABSTRACT FROM AUTHOR]

Abstract Mid-level element based representations have been proven to be very effective for visual recognition. This paper presents a method to discover discriminative mid-level visual elements based on deep Convolutional Neural Networks (CNNs). We present a part-level CNN architecture, namely Part-based CNN (P-CNN), which acts as a role of encoding module in a part-based representation model. The P-CNN can be attached at arbitrary layer of a pre-trained CNN and be trained using image-level labels. The training of P-CNN essentially corresponds to the optimization and selection of discriminative mid-level visual elements. For an input image, the output of P-CNN is naturally the part-based coding and can be directly used for image recognition. By applying P-CNN to multiple layers of a pre-trained CNN, more diverse visual elements can be obtained for visual recognitions. We validate the proposed P-CNN on several visual recognition tasks, including scene categorization, action classification and multi-label object recognition. Extensive experiments demonstrate the competitive performance of P-CNN in comparison with state-of-the-arts. [ABSTRACT FROM AUTHOR]

Category-independent region proposals have been utilized for salient objects detection in recent works. However, these works may fail when the extracted proposals have poor overlap with salient objects. In this paper, we demonstrate segment-level saliency prediction can provide these methods with complementary information to improve detection results. In addition, classification loss (i.e., softmax) can distinguish positive samples from negative ones and similarity loss (i.e., triplet) can enlarge the contrast difference between samples with different class labels. We propose a joint optimization of the two losses to further promote the performance. Finally, a multi-layer cellular automata model is incorporated to generate the final saliency map with fine shape boundary and object-level highlighting. The proposed method has achieved state-of-the-art results on four benchmark datasets. [ABSTRACT FROM AUTHOR]

Structural damage detection has been an interdisciplinary area of interest for various engineering fields. While the available damage detection methods have been in the process of adapting machine learning concepts, most machine learning based methods extract “hand-crafted” features which are fixed and manually selected in advance. Their performance varies significantly among various patterns of data depending on the particular structure under analysis. Convolutional neural networks (CNNs), on the other hand, can fuse and simultaneously optimize two major sets of an assessment task (feature extraction and classification) into a single learning block during the training phase. This ability not only provides an improved classification performance but also yields a superior computational efficiency. 1D CNNs have recently achieved state-of-the-art performance in vibration-based structural damage detection; however, it has been reported that the training of the CNNs requires significant amount of measurements especially in large structures. In order to overcome this limitation, this paper presents an enhanced CNN-based approach that requires only two measurement sets regardless of the size of the structure. This approach is verified using the experimental data of the Phase II benchmark problem of structural health monitoring which had been introduced by IASC-ASCE Structural Health Monitoring Task Group. As a result, it is shown that the enhanced CNN-based approach successfully estimated the actual amount of damage for the nine damage scenarios of the benchmark study. [ABSTRACT FROM AUTHOR]

We propose in this paper a new framework for 3D shape retrieval using queries of different modalities, which can include 3D models, images and sketches. The main scientific challenge is that different modalities have different representations and thus lie in different spaces. Moreover, the features that can be extracted from 2D images or 2D sketches are often different from those that can be computed from 3D models. Our solution is a new method based on Convolutional Neural Networks (CNN) that embeds all these entities into a common space. We propose a novel 3D shape descriptor based on local CNN features encoded using vectors of locally aggregated descriptors instead of conventional global CNN. Using a kernel function computed from 3D shape similarity, we build a target space in which wild images and sketches can be projected via two different CNNs. With this construction, matching can be performed in the common target space between same entities (sketch–sketch, image–image and 3D shape–3D shape) and more importantly across different entities (sketch-image, sketch-3D shape and image-3D shape). We demonstrate the performance of the proposed framework using different benchmarks including large scale SHREC 3D datasets. [ABSTRACT FROM AUTHOR]

This paper presents a learned feature based framework for both outdoor and indoor scene labeling. This framework is combined with a discriminative feature learning process to produce the posteriors of every pixel and a novel strategy to improve the global label consistency of a scene. First, we use Convolutional Neural Networks (ConvNets) to learn the most relevant features of a scene at the multi-scale superpixel level. The effect of both trained and general ConvNets features for our scene labeling framework are investigated. Then, based on the predicted posteriors from the learned features, we propose an algorithm called Region Consistency Activation (RCA) to iteratively improve the global label consistency at different levels of the Ultrametric Contour Map (UCM). In addition, we propose a strategy to make the hyper-parameters of RCA adaptive to the test images, which results in a better generalization ability compared with the hyper-parameters tuning based RCA. Our scene labeling framework were rigorously tested on three popular scene labeling datasets: Stanford Background, SIFT Flow and NYU-Depth V2. Experiments show that our proposed method consistently produces better accuracy and visual consistency compared with the state-of-the-art methods for both outdoor and indoor scenes. [ABSTRACT FROM AUTHOR]

Coming with the emerging of depth sensors link Microsoft Kinect, human hand gesture recognition has received ever increasing research interests recently. A successful gesture recognition system has usually heavily relied on having a good feature representation of data, which is expected to be task-dependent as well as coping with the challenges and opportunities induced by depth sensor. In this paper, a feature learning approach based on sparse auto-encoder (SAE) and principle component analysis is proposed for recognizing human actions, i.e. finger-spelling or sign language, for RGB-D inputs. The proposed model of feature learning is consisted of two components: First, features are learned respectively from the RGB and depth channels, using sparse auto-encoder with convolutional neural networks. Second, the learned features from both channels is concatenated and fed into a multiple layer PCA to get the final feature. Experimental results on American sign language (ASL) dataset demonstrate that the proposed feature learning model is significantly effective, which improves the recognition rate from 75% to 99.05% and outperforms the state-of-the-art. [ABSTRACT FROM AUTHOR]