• Neonatal brain segmentation in magnetic resonance (MR) is a challenging problem due to poor image quality and low contrast between white and gray matter regions. Most existing approaches for this problem are based on multi-atlas label fusion strategies, which are time-consuming and sensitive to registration errors. As alternative to these methods, we propose a hyper-densely connected 3D convolutional neural network that employs MR-T1 and T2 images as input, which are processed independently in two separated paths. An important difference with previous densely connected networks is the use of direct connections between layers from the same and different paths. Adopting such dense connectivity helps the learning process by including deep supervision and improving gradient flow. We evaluated our approach on data from the MICCAI Grand Challenge on 6-month infant Brain MRI Segmentation (iSEG), obtaining very competitive results. Among 21 teams, our approach ranked first or second in most metrics, translating into a state-of-the-art performance.
  • Recently, dense connections have attracted substantial attention in computer vision because they facilitate gradient flow and implicit deep supervision during training. Particularly, DenseNet, which connects each layer to every other layer in a feed-forward fashion, has shown impressive performances in natural image classification tasks. We propose HyperDenseNet, a 3D fully convolutional neural network that extends the definition of dense connectivity to multi-modal segmentation problems. Each imaging modality has a path, and dense connections occur not only between the pairs of layers within the same path, but also between those across different paths. This contrasts with the existing multi-modal CNN approaches, in which modeling several modalities relies entirely on a single joint layer (or level of abstraction) for fusion, typically either at the input or at the output of the network. Therefore, the proposed network has total freedom to learn more complex combinations between the modalities, within and in-between all the levels of abstraction, which increases significantly the learning representation. We report extensive evaluations over two different and highly competitive multi-modal brain tissue segmentation challenges, iSEG 2017 and MRBrainS 2013, with the former focusing on 6-month infant data and the latter on adult images. HyperDenseNet yielded significant improvements over many state-of-the-art segmentation networks, ranking at the top on both benchmarks. We further provide a comprehensive experimental analysis of features re-use, which confirms the importance of hyper-dense connections in multi-modal representation learning. Our code is publicly available at https://www.github.com/josedolz/HyperDenseNet.
  • Although influence maximization problem has been extensively studied over the past ten years, majority of existing work adopt one of the following models: \emph{full-feedback model} or \emph{zero-feedback model}. In the zero-feedback model, we have to commit the seed users all at once in advance, this strategy is also known as non-adaptive policy. In the full-feedback model, we select one seed at a time and wait until the diffusion completes, before selecting the next seed. Full-feedback model has better performance but potentially huge delay, zero-feedback model has zero delay but poorer performance since it does not utilize the observation that may be made during the seeding process. To fill the gap between these two models, we propose \emph{Partial-feedback Model}, which allows us to select a seed at any intermediate stage. We develop two novel greedy policies that, for the first time, achieve bounded approximation ratios under both uniform and non-uniform cost settings.
  • Purpose: To explore the usability and normal T1rho value of liver parenchyma with a novel single breathhold black blood single shot fast spin echo acquisition based liver imaging sequence. Materials and Methods: In total 19 health subjects (10 males, 9 females; mean age: 37.4 yrs; range: 23-54 yrs) participated in the study. 11 subjects had liver scanned twice in the same session to access scan-rescan repeatability. 12 subjects had liver scanned twice in two sessions with 7-10 days' interval to access scan-rescan reproducibility. MR was performed with a 3.0 T scanner with dual transmitter. The MR sequence allows simultaneous acquisition of 4 spin lock times (TSLs: 0ms, 10 ms, 30 ms, 50ms) in 10 second. Inherent black blood effect of fast spin echo and double inversion recovery were utilized to achieve blood signal suppression. Results: The technique demonstrated good image quality and minimal artifacts. For liver parenchyma, Bland-Altman plot showed the scan-rescan repeatability mean difference was 0.025 ms (95% limits of agreement: -1.163 to 1.213 ms), and intraclass correlation coefficient (ICC) was 0.977. The scan-rescan reproducibility mean difference was -0.075 ms (95% limits of agreement: -3.280 to 3.310 ms), and ICC was 0.820 which is better than the ICC of 0.764 of a previous bright blood multi-breath hold gradient echo acquisition technique. The liver T1rho value was 39.9 +/- 2.4 ms (range: 36.1 - 44.2 ms), which is lower than the value of 42.8=/-2.1 ms acquired with the previous bright blood technique. Conclusion: This study validated the application of a single breathhold black blood single shot fast spin echo acquisition based for human liver T1rho imaging. The lower liver parenchyma T1rho value and higher scan rescan reproducibility may improve of the sensitivity of this technique.
  • In this paper, we investigate the discount allocation problem in social networks. It has been reported that 40\% of consumers will share an email offer with their friend and 28\% of consumers will share deals via social media platforms. What does this mean for a business? Essentially discounts should not just be treated as short term solutions to attract individual customer, instead, allocating discounts to a small fraction of users (called seed users) may trigger a large cascade in a social network. This motivates us to study the influence maximization discount allocation problem: given a social network and budget, we need to decide to which initial set users should offer the discounts, and how much should the discounts be worth. Our goal is to maximize the number of customers who finally adopt the target product. We investigate this problem under both non-adaptive and adaptive settings. In the first setting, we have to commit the set of seed users and corresponding discounts all at once in advance. In the latter case, the decision process is performed in a sequential manner, and each seed user that is picked provides the feedback on the discount, or, in other words, reveals whether or not she will adopt the discount. We propose a simple greedy policy with an approximation ratio of $\frac{1}{2}(1 - 1/e)$ in non-adaptive setting. For the significantly more complex adaptive setting, we propose an adaptive greedy policy with bounded approximation ratio in terms of expected utility.
  • This paper first present a new general completely perturbed compressed sensing (CS) model y=(A+E)(x+u)+e,called \emph{noise folding based on general completely perturbed CS system},~where $y\in R^m,~u \in R^m,~u\neq 0,~e\in R^m$, $A\in R^{m\times n},~m\ll n$,~$E\in R^{m\times n}$ with incorporating general nonzero perturbation E to sensing matrix A and noise u into signal x simultaneously based on the standard CS model y=Ax+e.~Our constructions mainly will whiten the new proposed CS model and explore into RIP,~coherence for A+E of the new CS model after being whitened.
  • Pairing symmetries of iron-based superconductors are investigated systematically in a five-orbital model within the different regions of interaction parameters by functional renormalization group(FRG). Even for a fixed Fermi surface with both hole and electron pockets, it is found that depending on interaction parameters, a variety of pairing symmetries, including two types of $d$-wave and two types of $s$-wave pairing symmetries, can emerge. Only the $d_{x^2-y^2}$ and the $s\pm$ waves are robustly supported by the nearest-neighbor(NN) intra-orbital $J_1$ and the next-nearest-neighbor(NNN) intra-orbital $J_2$ antiferromagnetic(AFM) exchange couplings respectively. This study suggests that the accurate initial input of interaction parameters are essential to make FRG an useful method to determine the leading channel of superconducting instability.
  • As a novel technology, cloud computing attracts more and more people including technology enthusiasts and malicious users. Different from the classical network architecture, cloud environment has many its own features which make the traditional defense mechanism invalid. To make the network more robust against a malicious attack, we introduce a new method to mitigate this risk efficiently and systematically. In this paper, we first propose a coupled networks model which adequately considers the interactions between physical layer and virtual layer in a practical cloud computing environment. Based on this new model and our systematical method, we show that with the addition of protection of some specific nodes in the network structure, the robustness of cloud computing's network can be significantly improved whereas their functionality remains unchanged. Our results demonstrate that our new method can effectively settle the hard problems which cloud computing now is facing without much cost.
  • Although topological considerations amongst multiple labels have been previously investigated in the context of continuous max-flow image segmentation, similar investigations have yet to be made about shape considerations in a general and extendable manner. This paper presents shape complexes for segmentation, which capture more complex shapes by combining multiple labels and super-labels constrained by geodesic star convexity. Shape complexes combine geodesic star convexity constraints with hierarchical label organization, which together allow for more complex shapes to be represented. This framework avoids the use of co-ordinate system warping techniques to convert shape constraints into topological constraints, which may be ambiguous or ill-defined for certain segmentation problems.
  • For interdependent networks with identity dependency map, percolation is exactly the same with that on a single network and follows a second-order phase transition, while for random dependency, percolation follows a first-order phase transition. In real networks, the dependency relations between networks are neither identical nor completely random. Thus in this paper, we study the influence of randomness for dependency maps on the robustness of interdependent lattice networks. We introduce approximate entropy($ApEn$) as the measure of randomness of the dependency maps. We find that there is critical $ApEn_c$ below which the percolation is continuous, but for larger $ApEn$, it is a first-order transition. With the increment of $ApEn$, the $p_c$ increases until $ApEn$ reaching ${ApEn}_c'$ and then remains almost constant. The time scale of the system shows rich properties as $ApEn$ increases. Our results uncover that randomness is one of the important factors that lead to cascading failures of spatially interdependent networks.
  • Under the assumption that the superconducting state belongs to a single irreducible representation of lattice symmetry, we argue that the pairing symmetry in all measured iron-based superconductors is universally consistent with the A_{1g} s-wave. The robust s-wave pairing throughout the different families of iron-based superconductors at different doping regions signals two fundamental principles behind high T_c superconducting mechanisms: (1) the correspondence principle: the short range magnetic exchange interactions and the Fermi surfaces act collaboratively to achieve high T_c superconductivity and determine pairing symmetries; (2) the magnetic selection pairing rule: the superconductivity is only induced by the magnetic exchange couplings from the superexchange mechanism through cation-anion-cation chemical bondings. These principles explain why the unconventional high T_c superconductivity appears to be such a rare but robust phenomena with its strict requirement on electronic environment. The robust s-wave pairing also reveals that the current standard effective models with only onsite interactions are not sufficient and a minimum microscopic model must include strong nearest neighbor repulsive interactions resulted from the d-d direct bondings to serves as a s-wave symmetry stabilizer. Finally, the sign distribution of the superconducting order parameters in the reciprocal space is simply a consequence of the form factors given by the leading short-range pairings. The sign change of superconducting order parameters on Fermi surfaces is not a necessary requirement in repulsive-interaction-driven high T_c mechanism. The results will guide us to search for new electronic structure that supports high T_c superconductivity.
  • Social advertising (or social promotion) is an effective approach that produces a significant cascade of adoption through influence in the online social networks. The goal of this work is to optimize the ad allocation from the platform's perspective. On the one hand, the platform would like to maximize revenue earned from each advertiser by exposing their ads to as many people as possible, one the other hand, the platform wants to reduce free-riding to ensure the truthfulness of the advertiser. To access this tradeoff, we adopt the concept of \emph{regret} \citep{viral2015social} to measure the performance of an ad allocation scheme. In particular, we study two social advertising problems: \emph{budgeted social advertising problem} and \emph{unconstrained social advertising problem}. In the first problem, we aim at selecting a set of seeds for each advertiser that minimizes the regret while setting budget constraints on the attention cost; in the second problem, we propose to optimize a linear combination of the regret and attention costs. We prove that both problems are NP-hard, and then develop a constant factor approximation algorithm for each problem.
  • We predict that the recently discovered quasi-one dimensional superconductors, A$_2$Cr$_3$As$_3$(A=K,Rb), possess strong frustrated magnetic fluctuations and are nearby a novel in-out co-planar magnetic ground state. The frustrated magnetism is very sensitive to c-axis lattice constant and can thus be suppressed by increasing pressure. Our results qualitatively explain strong non-Fermi liquid behaviors observed in the normal state of the superconductors as the intertwining between the magnetism and superconductivity can create a large quantum critical region in quasi-one dimensional systems and also suggest that the materials share similar phase diagrams and superconducting mechanism with other unconventional superconductors, such as cuprates and iron-based superconductors.
  • One issue limiting the adaption of large-scale multi-region segmentation is the sometimes prohibitive memory requirements. This is especially troubling considering advances in massively parallel computing and commercial graphics processing units because of their already limited memory compared to the current random access memory used in more traditional computation. To address this issue in the field of continuous max-flow segmentation, we have developed a \textit{pseudo-flow} framework using the theory of Bregman proximal projections and entropic distances which implicitly represents flow variables between labels and designated source and sink nodes. This reduces the memory requirements for max-flow segmentation by approximately 20\% for Potts models and approximately 30\% for hierarchical max-flow (HMF) and directed acyclic graph max-flow (DAGMF) models. This represents a great improvement in the state-of-the-art in max-flow segmentation, allowing for much larger problems to be addressed and accelerated using commercially available graphics processing hardware.
  • Recent angle resolved photoemission spectroscopy(ARPES) experiments have suggested that BiS$_2$ based superconductors are at very low electron doping. Using random phase approximation(RPA) and functional renormalization group(FRG) methods, we find that $g$-wave pairing symmetry belonging to A$_{2g}$ irreducible representation is dominant at electron doping $x<0.25$. The pairing symmetry is determined by inter-pocket nesting and orbital characters on the Fermi surfaces and is robust in a two-orbital model including both Hund's coupling $J$, and Hubbard-like Coulomb interactions $U$ and $U'$ with relatively small $J$ ($J\leq0.2U$). With the increasing electron doping, the g-wave state competes with both the s-wave $A_{1g}$ and d-wave $B_{2g}$ states and no pairing symmetry emerges dominantly.
  • The incorporation of region regularization into max-flow segmentation has traditionally focused on ordering and part-whole relationships. A side effect of the development of such models is that it constrained regularization only to those cases, rather than allowing for arbitrary region regularization. Directed Acyclic Graphical Max-Flow (DAGMF) segmentation overcomes these limitations by allowing for the algorithm designer to specify an arbitrary directed acyclic graph to structure a max-flow segmentation. This allows for individual 'parts' to be a member of multiple distinct 'wholes.'
  • Multi-region segmentation algorithms often have the onus of incorporating complex anatomical knowledge representing spatial or geometric relationships between objects, and general-purpose methods of addressing this knowledge in an optimization-based manner have thus been lacking. This paper presents Generalized Hierarchical Max-Flow (GHMF) segmentation, which captures simple anatomical part-whole relationships in the form of an unconstrained hierarchy. Regularization can then be applied to both parts and wholes independently, allowing for spatial grouping and clustering of labels in a globally optimal convex optimization framework. For the purposes of ready integration into a variety of segmentation tasks, the hierarchies can be presented in run-time, allowing for the segmentation problem to be readily specified and alternatives explored without undue programming effort or recompilation.
  • Optimization techniques have been widely used in deformable registration, allowing for the incorporation of similarity metrics with regularization mechanisms. These regularization mechanisms are designed to mitigate the effects of trivial solutions to ill-posed registration problems and to otherwise ensure the resulting deformation fields are well-behaved. This paper introduces a novel deformable registration algorithm, RANCOR, which uses iterative convexification to address deformable registration problems under total-variation regularization. Initial comparative results against four state-of-the-art registration algorithms are presented using the Internet Brain Segmentation Repository (IBSR) database.
  • Using functional renormalization group (FRG) approach, we analyze the $\eta$ pairing in a five d-orbital model for iron-pncitides. We find that the $\eta$ pairing between two hole pockets is a leading instability at low energy when the two hole pockets are close to a nesting condition. The result suggests that the system can spontaneously enter into a superconducting state with both normal pairing and $\eta$ paring and the $\eta$ pairing is driven by the inter-orbital antiferromagnetic exchange coupling while the normal pairing is driven by the intra-orbital antiferromagnetic ones. An effective model at low energy is proposed to account for the $\eta$ pairing instability.
  • The recently proposed Magnetic Resonance Fingerprinting (MRF) technique can simultaneously estimate multiple parameters through dictionary matching. It has promising potentials in a wide range of applications. However, MRF introduces errors due to undersampling during the data acquisition process and the limit of dictionary resolution. In this paper, we investigate the error source of MRF and propose the technologies of improving the quality of MRF with compressed sensing, error prediction by decision trees, and adaptive filtering. Experimental results support our observations and show significant improvement of the proposed technologies.
  • Ni coated SiO2 and SiO2 coated Ni composite particles with core-shell structures were designed, prepared and their microwave absorption properties were characterized. The comparison study of the shell effect on the effective electromagnetic parameters reveals that the effective permittivity/permeability was crucially determined by the percent and the intrinsic electromagnetic parameters of the component materials regardless of various core/shell structures. Both of the composite core-shell structures could have good microwave absorption properties. Investigation of the input impedance indicates that, good microwave absorption performances are a consequence of proper electromagnetic impedance matches when the effective electromagnetic parameters were modulated.
  • In this paper we present two numerical schemes of approximating solutions of backward doubly stochastic differential equations (BDSDEs for short). We give a method to discretize a BDSDE. And we also give the proof of the convergence of these two kinds of solutions for BDSDEs respectively. We give a sample of computation of BDSDEs.