arXiv subjects

Jian Yang

Publications and source records attributed to Jian Yang.

At least 739 records · Page 41Linked to original sources

arXiv2018

Mixed Link Networks

Basing on the analysis by revealing the equivalence of modern networks, we find that both ResNet and DenseNet are essentially derived from the same "dense topology", yet they only differ in the form of connection -- addition (dubbed "inner link") vs. concatenation (dubbed "outer link"). However, both two forms of connections have the superiority and insufficiency. To combine their advantages and avoid certain limitations on representation learning, we present a highly efficient and modularized Mixed Link Network (MixNet) which is equipped with flexible inner link and outer link modules. Consequently, ResNet, DenseNet and Dual Path Network (DPN) can be regarded as a special case of MixNet, respectively. Furthermore, we demonstrate that MixNets can achieve superior efficiency in parameter over the state-of-the-art architectures on many competitive datasets like CIFAR-10/100, SVHN and ImageNet.

cs.LG↗

arXiv2018

Densely Connected Bidirectional LSTM with Applications to Sentence Classification

Deep neural networks have recently been shown to achieve highly competitive performance in many computer vision tasks due to their abilities of exploring in a much larger hypothesis space. However, since most deep architectures like stacked RNNs tend to suffer from the vanishing-gradient and overfitting problems, their effects are still understudied in many NLP tasks. Inspired by this, we propose a novel multi-layer RNN model called densely connected bidirectional long short-term memory (DC-Bi-LSTM) in this paper, which essentially represents each layer by the concatenation of its hidden state and all preceding layers' hidden states, followed by recursively passing each layer's representation to all subsequent layers. We evaluate our proposed model on five benchmark datasets of sentence classification. DC-Bi-LSTM with depth up to 20 can be successfully trained and obtain significant improvements over the traditional Bi-LSTM with the same or even less parameters. Moreover, our model has promising performance compared with the state-of-the-art approaches.

cs.CL↗

arXiv2018

SESR: Single Image Super Resolution with Recursive Squeeze and Excitation Networks

Single image super resolution is a very important computer vision task, with a wide range of applications. In recent years, the depth of the super-resolution model has been constantly increasing, but with a small increase in performance, it has brought a huge amount of computation and memory consumption. In this work, in order to make the super resolution models more effective, we proposed a novel single image super resolution method via recursive squeeze and excitation networks (SESR). By introducing the squeeze and excitation module, our SESR can model the interdependencies and relationships between channels and that makes our model more efficiency. In addition, the recursive structure and progressive reconstruction method in our model minimized the layers and parameters and enabled SESR to simultaneously train multi-scale super resolution in a single model. After evaluating on four benchmark test sets, our model is proved to be above the state-of-the-art methods in terms of speed and accuracy.

cs.CV↗

arXiv2018

The Composite Particle-Hole Spinor of the Lowest Landau Level

We propose to form a two-component effective field theory from L = (L_ce + L_ch)/2, where L_ce is the Lagrangian of composite electrons with a Chern-Simons term, and L_ch is the particle-hole conjugate of L_ce - the Lagrangian of composite holes. In the theory, the two-component fermion field phi is a composite particle-hole spinor coupled to an emergent effective gauge field in the presence of a background electromagnetic field. The Chern-Simons terms for both the composite electrons and composite holes are exactly cancelled out, and a 1/2 pseudospin degree of freedom, which responses to the emergent gauge field the same way as the real spin to the electromagnetic field, emerges automatically. Furthermore, the composite particle-hole spinor theory has exactly the same form as the non-relativistic limit of the massless Dirac composite fermion theory after expanded to the four-component form and with a mass term added.

cond-mat.str-el↗

arXiv2018

Understanding the Disharmony between Dropout and Batch Normalization by Variance Shift

This paper first answers the question "why do the two most powerful techniques Dropout and Batch Normalization (BN) often lead to a worse performance when they are combined together?" in both theoretical and statistical aspects. Theoretically, we find that Dropout would shift the variance of a specific neural unit when we transfer the state of that network from train to test. However, BN would maintain its statistical variance, which is accumulated from the entire learning procedure, in the test phase. The inconsistency of that variance (we name this scheme as "variance shift") causes the unstable numerical behavior in inference that leads to more erroneous predictions finally, when applying Dropout before BN. Thorough experiments on DenseNet, ResNet, ResNeXt and Wide ResNet confirm our findings. According to the uncovered mechanism, we next explore several strategies that modifies Dropout and try to overcome the limitations of their combination by avoiding the variance shift risks.

cs.LG↗

arXiv2018

Dirac Composite Fermion - A Particle-Hole Spinor

The particle-hole (PH) symmetry at half-filled Landau level requires the relationship between the flux number N_phi and the particle number N on a sphere to be exactly N_phi - 2(N-1) = 1. The wave functions of composite fermions with 1/2 "orbital spin", which contributes to the shift "1" in the N_phi and N relationship, are proposed, shown to be PH symmetric, and validated with exact finite system results. It is shown the many-body composite electron and composite hole wave functions at half-filling can be formed from the two components of the same spinor wave function of a massless Dirac fermion at zero-magnetic field. It is further shown that away from half-filling, the many-body composite electron wave function at filling factor nu and its PH conjugated composite hole wave function at 1-nu can be formed from the two components of the very same spinor wave functions of a massless Dirac fermion at non-zero magnetic field. This relationship leads to the proposal of a very simple Dirac composite fermion effective field theory, where the two-component Dirac fermion field is a particle-hole spinor field coupled to the same emergent gauge field, with one field component describing the composite electrons and the other describing the PH conjugated composite holes. As such, the density of the Dirac spinor field is the density sum of the composite electron and hole field components, and therefore is equal to the degeneracy of the Lowest Landau level. On the other hand, the charge density coupled to the external magnetic field is the density difference between the composite electron and hole field components, and is therefore neutral at exactly half-filling. It is shown that the proposed particle-hole spinor effective field theory gives essentially the same electromagnetic responses as Son's Dirac composite fermion theory does.

cond-mat.str-el↗

arXiv2017

Irreducible Projective Representations and Their Physical Applications

An eigenfunction method is applied to reduce the regular projective representations (Reps) of finite groups to obtain their irreducible projective Reps. Anti-unitary groups are treated specially, where the decoupled factor systems and modified Schur's lemma are introduced. We discuss the applications of irreducible Reps in many-body physics. It is shown that in symmetry protected topological phases, geometric defects or symmetry defects may carry projective Rep of the symmetry group; while in symmetry enriched topological phases, intrinsic excitations (such as spinons or visons) may carry projective Rep of the symmetry group. We also discuss the applications of projective Reps in problems related to spectrum degeneracy, such as in search of models without sign problem in quantum Monte Carlo Simulations.

cond-mat.str-el↗

arXiv2017

Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow Removal

Understanding shadows from a single image spontaneously derives into two types of task in previous studies, containing shadow detection and shadow removal. In this paper, we present a multi-task perspective, which is not embraced by any existing work, to jointly learn both detection and removal in an end-to-end fashion that aims at enjoying the mutually improved benefits from each other. Our framework is based on a novel STacked Conditional Generative Adversarial Network (ST-CGAN), which is composed of two stacked CGANs, each with a generator and a discriminator. Specifically, a shadow image is fed into the first generator which produces a shadow detection mask. That shadow image, concatenated with its predicted mask, goes through the second generator in order to recover its shadow-free image consequently. In addition, the two corresponding discriminators are very likely to model higher level relationships and global scene characteristics for the detected shadow region and reconstruction via removing shadows, respectively. More importantly, for multi-task learning, our design of stacked paradigm provides a novel view which is notably different from the commonly used one as the multi-branch version. To fully evaluate the performance of our proposed framework, we construct the first large-scale benchmark with 1870 image triplets (shadow image, shadow mask image, and shadow-free image) under 135 scenes. Extensive experimental results consistently show the advantages of ST-CGAN over several representative state-of-the-art methods on two large-scale publicly available datasets and our newly released one.

cs.CV↗

arXiv2017

Unsupervised Multi-Domain Image Translation with Domain-Specific Encoders/Decoders

Unsupervised Image-to-Image Translation achieves spectacularly advanced developments nowadays. However, recent approaches mainly focus on one model with two domains, which may face heavy burdens with large cost of $O(n^2)$ training time and model parameters, under such a requirement that $n$ domains are freely transferred to each other in a general setting. To address this problem, we propose a novel and unified framework named Domain-Bank, which consists of a global shared auto-encoder and $n$ domain-specific encoders/decoders, assuming that a universal shared-latent sapce can be projected. Thus, we yield $O(n)$ complexity in model parameters along with a huge reduction of the time budgets. Besides the high efficiency, we show the comparable (or even better) image translation results over state-of-the-arts on various challenging unsupervised image translation tasks, including face image translation, fashion-clothes translation and painting style translation. We also apply the proposed framework to domain adaptation and achieve state-of-the-art performance on digit benchmark datasets. Further, thanks to the explicit representation of the domain-specific decoders as well as the universal shared-latent space, it also enables us to conduct incremental learning to add a new domain encoder/decoder. Linear combination of different domains' representations is also obtained by fusing the corresponding decoders.

cs.CV↗

arXiv2017

FSRNet: End-to-End Learning Face Super-Resolution with Facial Priors

Face Super-Resolution (SR) is a domain-specific super-resolution problem. The specific facial prior knowledge could be leveraged for better super-resolving face images. We present a novel deep end-to-end trainable Face Super-Resolution Network (FSRNet), which makes full use of the geometry prior, i.e., facial landmark heatmaps and parsing maps, to super-resolve very low-resolution (LR) face images without well-aligned requirement. Specifically, we first construct a coarse SR network to recover a coarse high-resolution (HR) image. Then, the coarse HR image is sent to two branches: a fine SR encoder and a prior information estimation network, which extracts the image features, and estimates landmark heatmaps/parsing maps respectively. Both image features and prior information are sent to a fine SR decoder to recover the HR image. To further generate realistic faces, we propose the Face Super-Resolution Generative Adversarial Network (FSRGAN) to incorporate the adversarial loss into FSRNet. Moreover, we introduce two related tasks, face alignment and parsing, as the new evaluation metrics for face SR, which address the inconsistency of classic metrics w.r.t. visual perception. Extensive benchmark experiments show that FSRNet and FSRGAN significantly outperforms state of the arts for very LR face SR, both quantitatively and qualitatively. Code will be made available upon publication.

cs.CV↗

arXiv2017

Action-Attending Graphic Neural Network

The motion analysis of human skeletons is crucial for human action recognition, which is one of the most active topics in computer vision. In this paper, we propose a fully end-to-end action-attending graphic neural network (A$^2$GNN) for skeleton-based action recognition, in which each irregular skeleton is structured as an undirected attribute graph. To extract high-level semantic representation from skeletons, we perform the local spectral graph filtering on the constructed attribute graphs like the standard image convolution operation. Considering not all joints are informative for action analysis, we design an action-attending layer to detect those salient action units (AUs) by adaptively weighting skeletal joints. Herein the filtering responses are parameterized into a weighting function irrelevant to the order of input nodes. To further encode continuous motion variations, the deep features learnt from skeletal graphs are gathered along consecutive temporal slices and then fed into a recurrent gated network. Finally, the spectral graph filtering, action-attending and recurrent temporal encoding are integrated together to jointly train for the sake of robust action recognition as well as the intelligibility of human actions. To evaluate our A$^2$GNN, we conduct extensive experiments on four benchmark skeleton-based action datasets, including the large-scale challenging NTU RGB+D dataset. The experimental results demonstrate that our network achieves the state-of-the-art performances.

cs.CV↗

arXiv2017

Bufferless transmission in complex networks

Complex bufferless networks such as on-chip networks and optical burst switching networks haven't been paid enough attention in network science. In complex bufferless networks, the store and forward mechanism is not applicable, since the network nodes are not allowed to buffer data packets. In this paper, we study the data transmission process in complex bufferless networks from the perspective of network science. Specifically, we use the Price model to generate the underlying network topological structures. We propose a delivery queue based deflection mechanism, which accompanies the efficient routing protocol, to transmit data packets in bufferless networks. We investigate the average deflection times, packets loss rate, average arrival time, and how the network topological structure and some other factors affect these transmission performances. Our work provides some clues for the architecture and routing design of bufferless networks.

cs.NI↗

arXiv2017

MemNet: A Persistent Memory Network for Image Restoration

Recently, very deep convolutional neural networks (CNNs) have been attracting considerable attention in image restoration. However, as the depth grows, the long-term dependency problem is rarely realized for these very deep models, which results in the prior states/layers having little influence on the subsequent ones. Motivated by the fact that human thoughts have persistency, we propose a very deep persistent memory network (MemNet) that introduces a memory block, consisting of a recursive unit and a gate unit, to explicitly mine persistent memory through an adaptive learning process. The recursive unit learns multi-level representations of the current state under different receptive fields. The representations and the outputs from the previous memory blocks are concatenated and sent to the gate unit, which adaptively controls how much of the previous states should be reserved, and decides how much of the current state should be stored. We apply MemNet to three image restoration tasks, i.e., image denosing, super-resolution and JPEG deblocking. Comprehensive experiments demonstrate the necessity of the MemNet and its unanimous superiority on all three tasks over the state of the arts. Code is available at https://github.com/tyshiwo/MemNet.

cs.CV↗

arXiv2017

Spectral Filter Tracking

Visual object tracking is a challenging computer vision task with numerous real-world applications. Here we propose a simple but efficient Spectral Filter Tracking (SFT)method. To characterize rotational and translation invariance of tracking targets, the candidate image region is models as a pixelwise grid graph. Instead of the conventional graph matching, we convert the tracking into a plain least square regression problem to estimate the best center coordinate of the target. But different from the holistic regression of correlation filter based methods, SFT can operate on localized surrounding regions of each pixel (i.e.,vertex) by using spectral graph filters, which thus is more robust to resist local variations and cluttered background.To bypass the eigenvalue decomposition problem of the graph Laplacian matrix L, we parameterize spectral graph filters as the polynomial of L by spectral graph theory, in which L k exactly encodes a k-hop local neighborhood of each vertex. Finally, the filter parameters (i.e., polynomial coefficients) as well as feature projecting functions are jointly integrated into the regression model.

cs.CV↗

arXiv2017

Discriminative Block-Diagonal Representation Learning for Image Recognition

Existing block-diagonal representation researches mainly focuses on casting block-diagonal regularization on training data, while only little attention is dedicated to concurrently learning both block-diagonal representations of training and test data. In this paper, we propose a discriminative block-diagonal low-rank representation (BDLRR) method for recognition. In particular, the elaborate BDLRR is formulated as a joint optimization problem of shrinking the unfavorable representation from off-block-diagonal elements and strengthening the compact block-diagonal representation under the semi-supervised framework of low-rank representation. To this end, we first impose penalty constraints on the negative representation to eliminate the correlation between different classes such that the incoherence criterion of the extra-class representation is boosted. Moreover, a constructed subspace model is developed to enhance the self-expressive power of training samples and further build the representation bridge between the training and test samples, such that the coherence of the learned intra-class representation is consistently heightened. Finally, the resulting optimization problem is solved elegantly by employing an alternative optimization strategy, and a simple recognition algorithm on the learned representation is utilized for final prediction. Extensive experimental results demonstrate that the proposed method achieves superb recognition results on four face image datasets, three character datasets, and the fifteen scene multi-categories dataset. It not only shows superior potential on image recognition but also outperforms state-of-the-art methods.

cs.CV↗

arXiv2017

Adversarial PoseNet: A Structure-aware Convolutional Network for Human Pose Estimation

For human pose estimation in monocular images, joint occlusions and overlapping upon human bodies often result in deviated pose predictions. Under these circumstances, biologically implausible pose predictions may be produced. In contrast, human vision is able to predict poses by exploiting geometric constraints of joint inter-connectivity. To address the problem by incorporating priors about the structure of human bodies, we propose a novel structure-aware convolutional network to implicitly take such priors into account during training of the deep network. Explicit learning of such constraints is typically challenging. Instead, we design discriminators to distinguish the real poses from the fake ones (such as biologically implausible ones). If the pose generator (G) generates results that the discriminator fails to distinguish from real ones, the network successfully learns the priors.

cs.CV↗

arXiv2017

Game-theoretic Modeling of Players' Ambiguities on External Factors

We propose a game-theoretic framework that incorporates both incomplete information and general ambiguity attitudes on factors external to all players. Our starting point is players' preferences on payoff-distribution vectors, essentially mappings from states of the world to distributions of payoffs to be received by players. There are two ways in which equilibria for this preference game can be defined. When the preferences possess ever more features, we can gradually add ever more structures to the game. These include real-valued utility-like functions over payoff-distribution vectors, sets of probabilistic priors over states of the world, and eventually the traditional expected-utility framework involving one single prior. We establish equilibrium existence results, show the upper hemi-continuity of equilibrium sets over changing ambiguity attitudes, and uncover relations between the two versions of equilibria. Some attention is paid to the enterprising game, in which players exhibit ambiguity seeking attitudes while betting optimistically on the favorable resolution of ambiguities. The two solution concepts are unified at this game's pure equilibria, whose existence is guaranteed when strategic complementarities are present. The current framework can be applied to settings like auctions involving ambiguity on competitors' assessments of item worths.

econ.GN↗

arXiv2017

Analysis of Markovian Competitive Situations using Nonatomic Games

For dynamic situations where the evolution of a player's state is influenced by his own action as well as other players' states and actions, we show that equilibria derived for nonatomic games (NGs) can be used by their large finite counterparts to achieve near-equilibrium performances. We focus on the case with quite general spaces but also with independently generated shocks driving random actions and state transitions. The NG equilibria we consider are random state-to-action maps that pay no attention to players' external environments. They are adoptable by a variety of real situations where awareness of other players' states can be anywhere between full and non-existent. Transient results here also form the basis of a link between an NG's stationary equilibrium (SE) and good stationary profiles for large finite games.

econ.GN↗