Semi-supervised classification and betweenness computation on large, sparse, directed graphs

This work addresses graph-based semi-supervised classification and betweenness computation in large, sparse, networks (several millions of nodes). The objective of semi-supervised classification is to assign a label to unlabeled nodes using the whole topology of the graph and the labeling at our disposal. Two approaches are developed to avoid explicit computation of pairwise proximity between the nodes of the graph, which would be impractical for graphs containing millions of nodes. The first approach directly computes, for each class, the sum of the similarities between the nodes to classify and the labeled nodes of the class, as suggested initially in [1] and [2]. Along this approach, two algorithms exploiting different state-of-the-art kernels on a graph are developed. The same strategy can also be used in order to compute a betweenness measure. The second approach works on a trellis structure built from biased random walks on the graph, extending an idea introduced in [3]. These random walks allow to define a biased bounded betweenness for the nodes of interest, defined separately for each class. All the proposed algorithms have a linear computing time in the number of edges while providing good results, and hence are applicable to large sparse networks. They are empirically validated on medium-size standard data sets and are shown to be competitive with state-of-the-art techniques. Finally, we processed a novel data set, which is made available for benchmarking, for multi-class classification in a large network: the U.S. patents citation network containing 3M nodes (of six different classes) and 38M edges. The three proposed algorithms achieve competitive results (around 85% classification rate) on this large network-they classify the unlabeled nodes within a few minutes on a standard workstation.     

基于证据理论的纠错输出编码解决多类分类问题

针对多类分类问题,利用纠错输出编码作为分解框架,把多类问题转化为多个二类问题加以解决;同时提出一种基于证据理论的解码策略,把每一个二分器的输出作为证据之一进行融合,并讨论在两种编码类型(二元和三元编码矩阵)下证据融合的不同策略.通过实验分别对UCI数据集和3种一维距离像数据集进行测试,并与几种经典的解码方法进行比较,验证了所提出的方法能有效提高纠错输出编码特别是三元编码矩阵的分类正确率.     

3D-based Deep Convolutional Neural Network for action recognition with depth sequences

Traditional algorithms to design hand-crafted features for action recognition have been a hot research area in the last decade. Compared to RGB video, depth sequence is more insensitive to lighting changes and more discriminative due to its capability to catch geometric information of object. Unlike many existing methods for action recognition which depend on well-designed features, this paper studies deep learning-based action recognition using depth sequences and the corresponding skeleton joint information. Firstly, we construct a 3D-based Deep Convolutional Neural Network (3D²CNN) to directly learn spatio-temporal features from raw depth sequences, then compute a joint based feature vector named JointVector for each sequence by taking into account the simple position and angle information between skeleton joints. Finally, support vector machine (SVM) classification results from 3D²CNN learned features and JointVector are fused to take action recognition. Experimental results demonstrate that our method can learn feature representation which is time-invariant and viewpoint-invariant from depth sequences. The proposed method achieves comparable results to the state-of-the-art methods on the UTKinect-Action3D dataset and achieves superior performance in comparison to baseline methods on the MSR-Action3D dataset. We further investigate the generalization of the trained model by transferring the learned features from one dataset (MSR-Action3D) to another dataset (UTKinect-Action3D) without retraining and obtain very promising classification accuracy.     

Mixed graph convolution and residual transformation network for skeleton-based action recognition

Action recognition based on a human skeleton is an extremely challenging research problem. The temporal information contained in the human skeleton is more difficult to extract than the spatial information. Many researchers focus on graph convolution networks and apply them to action recognition. In this study, an action recognition method based on a two-stream network called RNXt-GCN is proposed on the basis of the Spatial-Temporal Graph Convolutional Network (ST-GCN). The human skeleton is converted first into a spatial-temporal graph and a SkeleMotion image which are input into ST-GCN and ResNeXt, respectively, for performing the spatial-temporal convolution. The convolved features are then fused. The proposed method models the temporal information in action from the amplitude and direction of the action and addresses the shortcomings of isolated temporal information in the ST-GCN. The experiments are comprehensively performed on the four datasets: 1) UTD-MHAD, 2) Northwestern-UCLA, 3) NTU RGB-D 60, and 4) NTU RGB-D 120. The proposed model shows very competitive results compared with other models in our experiments. On the experiments of NTU RGB?+?D 120 dataset, our proposed model outperforms those of the state-of-the-art two-stream models.

     

Randomness, Stochasticity, and Approximations

Polynomial time unsafe approximations for intractable sets were introduced by Meyer and Paterson [9] and Yesha [19], respectively. The question of which sets have optimal unsafe approximations has been investigated extensively; see, e.g., [1], [5], [15], and [17]. Recently, Wang [15], [17] showed that polynomial time random sets are neither optimally unsafe approximable nor Δ -levelable. In this paper we show that: (1) There exists a polynomial time stochastic set in the exponential time complexity class which has an optimal unsafe approximation. (2) There exists a polynomial time stochastic set in the exponential time complexity class which is Δ -levelable. The above two results answer a question asked by Ambos-Spies and Lutz [2]: What kind of natural complexity property can be characterized by p -randomness but not by p -stochasticity? Our above results also extend Ville's [13] historical result. The proof of our first result shows that, for Ville's stochastic sequence, we can find an optimal prediction function f such that we will never lose our own money betting according to f (except the money we have earned), that is to say, if at the beginning we have only $1 and we always bet $1 that the next selected bit is 1 , then we always have enough money to bet on the next bit. Our second result shows that there is a stochastic sequence for which there is a betting strategy f such that we will never lose our own money betting according to f (except the money we have earned), but there is no such optimal betting strategy. That is to say, for any such betting strategy, we can find another betting strategy which could be used to make money more quickly. Received May 1997, and in final form September 1998.     

3D shape recognition based on multi-modal information fusion

The classification and retrieval of 3D models have been widely used in the field of multimedia and computer vision. With the rapid development of computer graphics, different algorithms corresponding to different representations of 3D models have achieved the best performance. The advances in deep learning also encourage various deep models for 3D feature representation. For multi-view, point cloud, and PANORAMA-view, different models have shown significant performance on 3D shape classification. However, There’s not a way to consider utilizing the fusion information of multi-modal for 3D shape classification. In our opinion, We propose a novel multi-modal information fusion method for 3D shape classification, which can fully utilize the advantage of different modal to predict the label of class. More specifically, the proposed can effectively fuse more modal information. it is easy to utilize in other similar applications. We have evaluated our framework on the popular dataset ModelNet40 for the classification task on 3D shape. Series experimental results and comparisons with state-of-the-art methods demonstrate the validity of our approach.

     

Faster CNN-based vehicle detection and counting strategy for fixed camera scenes

Automatic detection and counting of vehicles in a video is a challenging task and has become a key application area of traffic monitoring and management. In this paper, an efficient real-time approach for the detection and counting of moving vehicles is presented based on YOLOv2 and features point motion analysis. The work is based on synchronous vehicle features detection and tracking to achieve accurate counting results. The proposed strategy works in two phases; the first one is vehicle detection and the second is the counting of moving vehicles. Different convolutional neural networks including pixel by pixel classification networks and regression networks are investigated to improve the detection and counting decisions. For initial object detection, we have utilized state-of-the-art faster deep learning object detection algorithm YOLOv2 before refining them using K-means clustering and KLT tracker. Then an efficient approach is introduced using temporal information of the detection and tracking feature points between the framesets to assign each vehicle label with their corresponding trajectories and truly counted it. Experimental results on twelve challenging videos have shown that the proposed scheme generally outperforms state-of-the-art strategies. Moreover, the proposed approach using YOLOv2 increases the average time performance for the twelve tested sequences by 93.4% and 98.9% from 1.24 frames per second achieved using Faster Region-based Convolutional Neural Network (F R-CNN ) and 0.19 frames per second achieved using the background subtraction based CNN approach (BS-CNN ), respectively to 18.7 frames per second.

     

Video multimodal emotion recognition based on Bi-GRU and attention fusion

A video multimodal emotion recognition method based on Bi-GRU and attention fusion is proposed in this paper. Bidirectional gated recurrent unit (Bi-GRU) is applied to improve the accuracy of emotion recognition in time contexts. A new network initialization method is proposed and applied to the network model, which can further improve the video emotion recognition accuracy of the time-contextual learning. To overcome the weight consistency of each modality in multimodal fusion, a video multimodal emotion recognition method based on attention fusion network is proposed. The attention fusion network can calculate the attention distribution of each modality at each moment in real-time so that the network model can learn multimodal contextual information in real-time. The experimental results show that the proposed method can improve the accuracy of emotion recognition in three single modalities of textual, visual, and audio, meanwhile improve the accuracy of video multimodal emotion recognition. The proposed method outperforms the existing state-of-the-art methods for multimodal emotion recognition in sentiment classification and sentiment regression.

     

Relationship between tau,neuroinflammation and atrophy in Alzheimer's disease: The NIMROD study

In addition to beta-amyloid accumulation, misfolded tau and activated microglia are also present in Alzheimer's disease (AD). It is important to study the relationship amongst these pathologies in vivo and their effects on the cognitive deficits for developing effective trails and future therapeutic or preventive strategies for AD. To investigate the relationships amongst different pathologies in AD, in particular how they interact resulting in cognitive impairments, we conducted a study of sixty-six subjects (15 AD, 24 Mild Cognitive Impairment (MCI) and 27 similarly aged healthy controls), who underwent standardised clinical and neuropsychological assessments followed by dynamic PET using [¹⁸F]AV1451 (tau) and [¹¹C]PK11195 (activated microglia) and multimodal 3T MRI. MCI patients also underwent [¹¹C]PIB (beta-amyloid) PET. We compared regional PET binding and grey matter atrophy amongst AD, amyloid positive MCI and controls, as well as their spatial distribution across different brain areas. We also applied a mediation analysis to infer the direct and indirect effects of tau, neuroinflammation and grey matter atrophy on cognitive functioning. We found increased [¹⁸F]AV1451 and [¹¹C]PK11195 binding as well as grey matter atrophy in AD, with a strong spatial overlap amongst these AD related biomarkers suggesting them interacting with each other. We demonstrated that both tau ([¹⁸F]AV1451) and neuroinflammation ([¹¹C]PK11195) have significant effects on cognition however their effects were fully mediated by grey matter atrophy. No mediation effect between tau and neuroinflammation were found with respect to cognition. In conclusion, grey matter atrophy not only spatially overlapped with tau and microglia activity in AD, but also mediate them in affecting cognitive impairments. The mediation analysis enabled data fusion across multiple imaging modalities (PET and MRI) and multiple PET tracers. Our results have significant implications for trials targeting tau and inflammation, and future therapeutic or preventive strategies for AD.     

A Faster FPT Algorithm for the Maximum Agreement Forest Problem

Given two unrooted, binary trees, T₁ and T₂, leaf labelled bijectively by a set of species L, the Maximum Agreement Forest (MAF) problem asks to find a minimum cardinality collection F = {t₁, ..., t_k} of phylogenetic trees where each element of F is a subtree of both T₁ and T₂, the elements of F are pairwise disjoint, and the leaf labels for the elements of F partition the leaf label set L. We give an efficient fixed-parameter tractable (FPT) algorithm for the MAF problem, significantly improving on an FPT algorithm given in [2]. Whereas the algorithm from [2] has a running time of O(k^3k) + p(|L|), our algorithm runs in time O(4^k · k⁵) + p(|L|), where k bounds the size of the agreement forest and p(·) is a low order polynomial.     

On closure properties of bounded two-sided error complexity classes

We show that if a complexity classC is closed downward under polynomial-time majority truth-table reductions ( _mtt ^p ), then practically every other polynomial closure property it enjoys is inherited by the corresponding bounded two-sided error class BP[C]. For instance, the Arthur-Merlin game class AM [B1] enjoys practically every closure property of NP. Our main lemma shows that, for any relativizable classD which meets two fairly transparent technical conditions, we haveC ^BP[C] BP[D ^C]. Among our applications, we simplify the proof by Toda [Tol], [To2] that the polynomial hierarchy PH is contained in BP[P]. We also show that relative to a random oracleR, PH ^R is properly contained in P ^R .The first author was supported in part by NSF Grant CCR-9011248 and the second author was supported in part by NSF Grant CCR-89011154.     

A fast algorithm to compute irreducible and primitive polynomials in finite fields

In this paper we present a method to computeall the irreducible and primitive polynomials of degreem over the finite fieldGF(q). Our method finds each new irreducible or primitive polynomial with a complexity ofO(m) arithmetic operations inGF(q). The best previously known methods [3], [10] use the Berlekamp-Massey algorithm [7] and they have a complexityO(m ²). We reach mis improvement taking into account a systolic implementation [2] of the extended Euclidean algorithm instead of using the Berlekamp-Massey algorithm.This work was supported in part by Spanish Grant CICYT TIC91-0472.     

J-SYSTEM RECONSTRUCT/ABILITY: A FORMALIZED STUDY

TheJ-system satisfies either J =P([,4], [.B], [C]); or J = [A], [B] [C]) or J = [A]O₁, [B] O₂ [C] Therefore, J-system reconstructability (I) is in motion, in accordance with the formalized law AOB?C; (ii) is in development and evolution, in accordance wiih f(A,B,C)→F(lA,m£,nC); and (iii) has qualitative change, in accordance with (δ₁A)O₇(δ ₂B)O₈ ( δ ₈C) = KδP. This paper discusses these laws in detail.     

Fast moving pedestrian detection based on motion segmentation and new motion features

The detection of moving pedestrians is of major importance for intelligent vehicles, since information about such persons and their tracks should be incorporated into reliable collision avoidance algorithms. In this paper, we propose a new approach to detect moving pedestrians aided by motion analysis. Our main contribution is to use motion information in two ways: on the one hand we localize blobs of moving objects for regions of interest (ROIs) selection by segmentation of an optical flow field in a pre-processing step, so as to significantly reduce the number of detection windows needed to be evaluated by a subsequent people classifier, resulting in a fast method suitable for real-time systems. On the other hand we designed a novel kind of features called Motion Self Difference (MSD) features as a complement to single image appearance features, e. g. Histograms of Oriented Gradients (HOG), to improve distinctness and thus classifier performance. Furthermore, we integrate our novel features in a two-layer classification scheme combining a HOG+Support Vector Machines (SVM) and a MSD+SVM detector. Experimental results on the Daimler mono moving pedestrian detection benchmark show that our approach obtains a log-average miss rate of 36 % in the FPPI range [10^?2,10⁰], which is a clear improvement with respect to the naive HOG+SVM approach and better than several other state-of-the-art detectors. Moreover, our approach also reduces runtime per frame by an order of magnitude.     

Ring models for delay-differential systems

This paper studies the algebraic structure of linear systems defined over R[λ], the ring of polynomials in λ with real coefficients. Natural definitions of controllability and observability are introduced and properties of R[λ]-transfer matrix realizations are discussed. It is shown that (A_n×n,D_n×m) is a controllable R[λ]-matrix pair if and only if for each set of polynomialsβ₁,β₂,…,β_n, in R[λ] there exists an R[λ] feedback matrixF such that detsI?A?BF]=∏i=1n(s+β_i). By regarding λ as a suitably defined delay operator, it is explained how this result might be applied to delay-differential systems in order to control dynamic response.     

基于视频分段的空时双通道卷积神经网络的行为识别

针对原始空时双通道卷积神经网络（CNN）模型对长时段复杂视频中行为识别率低的问题，提出了一种基于视频分段的空时双通道卷积神经网络的行为识别方法。首先将视频分成多个等长不重叠的分段，对每个分段随机采样得到代表视频静态特征的帧图像和代表运动特征的堆叠光流图像；然后将这两种图像分别输入到空域和时域卷积神经网络进行特征提取，再在两个通道分别融合各视频分段特征得到空域和时域的类别预测特征；最后集成双通道的预测特征得到视频行为识别结果。通过实验讨论了多种数据增强方法和迁移学习方案以解决训练样本不足导致的过拟合问题，分析了不同分段数、预训练网络、分段特征融合方案和双通道集成策略对行为识别性能的影响。实验结果显示所提模型在UCF101数据集上的行为识别准确率达到91.80%，比原始的双通道模型提高了3.8个百分点；同时在HMDB51数据集上的行为识别准确率也比原模型提高，达到61.39%，这表明所提模型能够更好地学习和表达长时段复杂视频中人体行为特征。     

Time-aware hybrid expertise retrieval system in community question answering services

This paper introduces a time-aware hybrid expertise retrieval (TaHER) system for community question answering (CQA) services. It comprises of a text-based part and a network-based part. The text-based part makes use of the textual and the temporal information associated with questions and answers. Moreover, it assesses the recent interests and the activities of answerers. For a given question, it determines the knowledge of each answerer and identify active answerers with adequate knowledge. The network-based part is composed of several period-dependent networks. It uses the relationships among the answerers along with temporal information. Next, it applies a link analysis technique on the networks to determine the time-aware authority of each answerer in the community. We, nonetheless, propose a fusion strategy for combining the offshoots of these two parts. Using 5 performance measures, TaHER system is compared with 20 state-of-the-art algorithms on 4 real-world datasets. According to our experiments, in 93.75% (375 out of 400) cases, the proposed approach outperforms the comparing approaches. We also experimentally validate the importance of each assumption used by us.

     

OLEDs based on an europium(IIIrpar; complex: {Tris(thenoyltrifluoroacetonate)[1,2,5]‐thiadiazolo[3,4‐f] [1,10]phenanthroline}europium(III)

Abstract— The contribution of radiative and non‐radiative processes to the electroluminescence emission of OLEDs based on Eu‐complex, {tris(thenoyltrifluoroacetone)[1,2,5]thiadiazolo[3,4‐f][1,10]phenanthroline} europium(III), [Eu(TTA)³TDZP], which acts as transporting and emitting layers, is investigated. The Eu‐complex presented an intense photoluminescence with high color purity in the red region, characteristic of the Eu(III) ⁵D⁰ → ⁷F² narrow line transition. However, when used in a double‐layered OLED its electroluminescence showed additional undesired broad bands, which can be attributed to the possible electrophosphorescence of the ligand and to an inefficient energy transfer from the organic ligand to the Eu(III). The characteristic narrow lines could be achieved using a co‐deposited active layer with the Eu‐complex acting as a dopant in a matrix comprised of 4,4’‐bis(carbazol‐9‐yl)biphenyl (CBP).     

Semantic classification of mobile robot locations through 2D laser scans

Previous learning-based methods that rely on 2D laser data to classify indoor robot locations into semantic classes were successful in distinguishing between rooms and corridors. However, the classification accuracy remained low for doorway locations. We propose a semantic place classification method that uses a rule-based doorway detection algorithm followed by a classification scheme that models training data through either K-means clustering or learning vector quantization. We conducted extensive experiments on the Freiburg 79 dataset and compared our method to previous semantic place classification algorithms. The doorway detection algorithm we propose significantly increases the classification accuracy for doorway locations as compared to the state-of-the-art performance. We applied our method, trained on the Freiburg 79 dataset, to Freiburg 52 and ESOGU datasets in order to demonstrate its generalization ability.

     

信息分离和质量引导的红外与可见光图像融合

目的 红外与可见光图像融合的目标是将红外图像与可见光图像的互补信息进行融合,增强源图像中的细节场景信息。然而现有的深度学习方法通常人为定义源图像中需要保留的特征,降低了热目标在融合图像中的显著性。此外,特征的多样性和难解释性限制了融合规则的发展,现有的融合规则难以对源图像的特征进行充分保留。针对这两个问题,本文提出了一种基于特有信息分离和质量引导的红外与可见光图像融合算法。方法 本文提出了基于特有信息分离和质量引导融合策略的红外与可见光图像融合算法。设计基于神经网络的特有信息分离以将源图像客观地分解为共有信息和特有信息,对分解出的两部分分别使用特定的融合策略;设计权重编码器以学习质量引导的融合策略,将衡量融合图像质量的指标应用于提升融合策略的性能,权重编码器依据提取的特有信息生成对应权重。结果 实验在公开数据集RoadScene上与6种领先的红外与可见光图像融合算法进行了对比。此外,基于质量引导的融合策略也与4种常见的融合策略进行了比较。定性结果表明,本文算法使融合图像具备更显著的热目标、更丰富的场景信息和更多的信息量。在熵、标准差、差异相关和、互信息及相关系数等指标上,相较于对比算法...