查看更多>>摘要：This paper develops Bayesian Compression for Dynamically Expandable Network (BCDEN), which can learn a compact model structure with preserving the accuracy in a continual learning scenarios. Dynamically Expandable Network (DEN) is efficiently trained by performing selective retraining, dynamically expands network capacity with only the necessary number of units, and effectively prevents semantic drift by duplicating and timestamping units in an online manner. Overcoming conventional DEN only giving point estimates, we providing the Bayesian inference under the principle framework. We validate our BCDEN on multiple public datasets under continual learning setting, on which it can outperform existing continual learning methods on a variety of tasks, and with the state-of-the-art compression results, while still maintaining comparable performance. (c) 2021 Elsevier Ltd. All rights reserved.
查看更多>>摘要：Recently, exploring features from different layers in fully convolutional networks (FCNs) has gained sub-stantial attention to capture context information for semantic segmentation. This paper presents a novel encoder-decoder architecture, called contextual ensemble network (CENet), for semantic segmentation, where the contextual cues are aggregated via densely usampling the convolutional features of deep layer to the shallow deconvolutional layers. The proposed CENet is trained in terms of end-to-end segmenta-tion to match the resolution of input image, and allows us to fully explore contextual features through ensemble of dense deconvolutions. We evaluate our CENet on two widely-used semantic segmentation datasets: PASCAL VOC 2012 and CityScapes. The experimental results demonstrate our CENet achieves superior performance with respect to recent state-of-the-art results. Furthermore, we also evaluate CENet on MS COCO dataset and ISBI 2012 dataset for the task of instance segmentation and biological segmen-tation, respectively. The experimental results show that CENet obtains promising results on these two datasets. (c) 2021 Elsevier Ltd. All rights reserved.
查看更多>>摘要：This work presents a new deep learning approach for keystroke biometrics based on a novel Distance Metric Learning method (DML). DML maps input data into a learned representation space that reveals a "semantic" structure based on distances. In this work, we propose a novel DML method specifically de-signed to address the challenges associated to free-text keystroke identification where the classes used in learning and inference are disjoint. The proposed SetMargin Loss (SM-L) extends traditional DML ap-proaches with a learning process guided by pairs of sets instead of pairs of samples, as done traditionally. The proposed learning strategy allows to enlarge inter-class distances while maintaining the intra-class structure of keystroke dynamics. We analyze the resulting representation space using the mathemati-cal problem known as Circle Packing, which provides neighbourhood structures with a theoretical max-imum inter-class distance. We finally prove experimentally the effectiveness of the proposed approach on a challenging task: keystroke biometric identification over a large set of 78,0 0 0 subjects. Our method achieves state-of-the-art accuracy on a comparison performed with the best existing approaches. (c) 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )
查看更多>>摘要：Conditional Random Fields (CRF) are frequently applied for labeling and segmenting sequence data. Morency et al. (2007) introduced hidden state variables in a labeled CRF structure in order to model the latent dynamics within class labels, thus improving the labeling performance. Such a model is known as Latent-Dynamic CRF (LDCRF). We present Factored LDCRF (FLDCRF), a structure that allows multiple latent dynamics of the class labels to interact with each other. Including such latent-dynamic interactions leads to improved labeling performance on single-label and multi-label sequence modeling experiments across two different datasets, viz., UCI gesture phase data and UCI opportunity data. FLDCRF outperforms all state-of-the-art sequence models, viz., CRF, LDCRF, L STM, L STM-CRF, Factorial CRF, Coupled CRF and a multi-label LSTM model across experiments in this paper. In addition, FLDCRF offers easier model selection and is more consistent across validation and test data than LSTM models. FLDCRF is also much faster to train compared to LSTM, even without a GPU. FLDCRF outshines the best LSTM model by-4% on a single-label task on the UCI gesture phase data and outperforms LSTM models by-2% on average on the multi-label sequence tagging experiment on the UCI opportunity data. (c) 2021 Elsevier Ltd. All rights reserved.
查看更多>>摘要：Despite the impressive performance of random forests (RF), its theoretical properties have not been thoroughly understood. In this paper, we propose a novel RF framework, dubbed multinomial random forest (MRF), to analyze its consistency and privacy-preservation . Instead of deterministic greedy split rule or with simple randomness, the MRF adopts two impurity-based multinomial distributions to randomly select a splitting feature and a splitting value, respectively. Theoretically, we prove the consistency of MRF and analyze its privacy-preservation within the framework of differential privacy. We also demonstrate with multiple datasets that its performance is on par with the standard RF. To the best of our knowledge, MRF is the first consistent RF variant that has comparable performance to the standard RF. The code is available at https://github.com/jiawangbai/Multinomial- Random-Forest . (c) 2021 Published by Elsevier Ltd.
Evangelista, Rodrigo C.Salvadeo, Denis H. P.Mascarenhas, Nelson D. A.
10页
查看更多>>摘要：Poisson noise is the main cause of degradation of many imaging modalities. However, many of the proposed methods for reducing noise in images lack a formal approach. Our work develops a new, general, formal and computationally efficient bayesian Poisson denoising algorithm, based on the Nonlocal Means framework and replacing the euclidean distance by stochastic distances, which are more appropriate for the denoising problem. It takes advantage of the conjugacy of Poisson and gamma distributions to obtain its computational efficiency. When dealing with low dose CT images, the algorithm operates on the sinogram, modeling the rates of the Poisson noise by the Gamma distribution. Based on the Bayesian formulation and the conjugacy property, the likelihood follows the Poisson distribution, while the a posteriori distribution is also described by the Gamma distribution. The derived algorithm is applied to simulated and real low-dose CT images and compared to several algorithms proposed in the literature, with competitive results. (c) 2021 Elsevier Ltd. All rights reserved.
Raj, SridharPrasad, Munaga V. N. K.Balakrishnan, Ramadoss
16页
查看更多>>摘要：Multi-camera video surveillance environment has a variety of emerging research problems among, which person re-identification is the premier one. Unsupervised person re-identification has been explored less in literature than the supervised approach. Images acquired from the video surveillance systems are unlabeled, which denotes that it is naturally an unsupervised learning problem. The state-of-the-art unsupervised methods seek external annotations support such as incorporating transfer learning techniques, partial labeling of train images, etc., which makes them not purely unsupervised and unsuitable for practical real-world surveillance settings. Identity mismatch happens due to the similar costumes and complex environmental factors. To resolve this issue, we introduce a new framework named Spatio-Temporal Association Rule based Deep Annotation-free Clustering (STAR-DAC) which incrementally clusters the unlabeled person re-identification images based on visual features and performs cluster fine-tuning through the mined spatio-temporal association rules. STAR formulations leveraged upto 75% of images for reliable sample selection through cluster fine-tuning. STAR based fine-tune algorithm aims to attain ground-truth labels of an unlabeled dataset and eliminate cluster outliers to stabilize the evaluation. Experiments are performed on image and video-based benchmark person re-identification datasets such as DukeMTMC re-ID, Market1501, MSMT17, CUHK03, GRID and Dukevideo re-ID, iLIDSVid, ViPer respectively. Experimental results clearly show that the proposed STAR-DAC framework outperforms the state-of-the-art methods in case of large scale datasets with multiple cameras. (c) 2021 Elsevier Ltd. All rights reserved.
查看更多>>摘要：Abundant data are essential for improving the performance of machine learning algorithms. Thus, if only limited data are available, data synthesis can be used to enlarge datasets. Data synthesis methods based on the covariance matrix are useful because of their fast data synthesis capabilities. However, artifi-cial datasets generated via classical techniques show statistical discrepancies when compared to original datasets. To address this problem, we developed a new data synthesis method that preserves the corre-lation (between features) observed in the original dataset. This preservation was realized by considering not only the correlation but also the random noises used in data synthesis process. This method was applied to various biosignals (i.e., electrocortiography, electromyogram, and electrocardiogram), wherein data points are insufficient. Several classifiers (i.e., convolutional neural network, support vector machine, and k-nearest neighbor) were used to verify that the classification accuracy can be improved by the pro-posed data synthesis method. (c) 2021 Elsevier Ltd. All rights reserved.
查看更多>>摘要：Lung cancer classification in screening computed tomography (CT) scans is one of the most crucial tasks for early detection of this disease. Many lives can be saved if we are able to accurately classify malignant/cancerous lung nodules. Consequently, several deep learning based models have been proposed recently to classify lung nodules as malignant or benign. Nevertheless, the large variation in the size and heterogeneous appearance of the nodules makes this task an extremely challenging one. We propose a new Progressive Growing Channel Attentive Non-Local (ProCAN) network for lung nodule classification. The proposed method addresses this challenge from three different aspects. First, we enrich the Non Local network by adding channel-wise attention capability to it. Second, we apply Curriculum Learning principles, whereby we first train our model on easy examples before hard ones. Third, as the classification task gets harder during the Curriculum learning, our model is progressively grown to increase its capability of handling the task at hand. We examined our proposed method on two different public datasets and compared its performance with state-of-the-art methods in the literature. The results show that the ProCAN model outperforms state-of-the-art methods and achieves an AUC of 98.05% and an accuracy of 95.28% on the LIDC-IDRI dataset. Moreover, we conducted extensive ablation studies to analyze the contribution and effects of each new component of our proposed method. (c) 2021 Elsevier Ltd. All rights reserved.
查看更多>>摘要：Multi-view clustering has become a hot yet challenging topic, due mainly to the independence of and information complementarity between different views. Although good results are achieved to a certain extent from typical methods including multi-view based k-means clustering, sparse cooperative representation clustering and subspace clustering, they still suffer from several drawbacks or limitations: (1) When each view is sparse decomposed, it still contains some hidden information for mining, such as the structure of samples, the intra-class similarity measure, and the inter-class diversity discrimination, etc. (2) Most of the existing multi-view methods only consider the local features within each view, but fail to effectively balance the importance of and combine information among different views in a diversified way. To tackle these issues, we propose a novel multi-view diversity learning model based on robust bilinear error decomposition (BED). The BED term with a low rank sparse constraint is an improved non negative matrix factorization (NMF), which is used to extract the hidden structure information in sparse decomposition and useful diversity discrimination information in error matrix. The preservation of local features and selection of important views are achieved by adaptive weighted manifold learning. Furthermore, the Hilbert Schmidt independence criterion is used as a diversity learning term for mutual learning and fusion among views. Finally, the proposed robust low-rank multi-view diversity learning spectral clustering method is evaluated and benchmarked with eight state-of-the-art methods. Experiments in six real datasets have fully validated the significantly improved accuracy and efficiency of the proposed methodology for effective clustering of multi-view images. (c) 2021 Elsevier Ltd. All rights reserved.
NSTL
Elsevier