The Experts below are selected from a list of 327 Experts worldwide ranked by ideXlab platform

Ming-hsuan Yang - One of the best experts on this subject based on the ideXlab platform.

  • Diversified Texture Synthesis with Feed-Forward Networks
    arXiv: Computer Vision and Pattern Recognition, 2017
    Co-Authors: Chen Fang, Jimei Yang, Zhaowen Wang, Ming-hsuan Yang
    Abstract:

    Recent progresses on deep discriminative and generative modeling have shown promising results on texture synthesis. However, existing feed-Forward based methods trade off generality for efficiency, which suffer from many issues, such as shortage of generality (i.e., build one network per texture), lack of diversity (i.e., always produce visually identical output) and suboptimality (i.e., generate less satisfying visual effects). In this work, we focus on solving these issues for improved texture synthesis. We propose a deep generative feed-Forward network which enables efficient synthesis of multiple textures within one single network and meaningful interpolation between them. Meanwhile, a suite of important techniques are introduced to achieve better convergence and diversity. With extensive experiments, we demonstrate the effectiveness of the proposed model and techniques for synthesizing a large number of textures and show its applications with the stylization.

  • CVPR - Diversified Texture Synthesis with Feed-Forward Networks
    2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017
    Co-Authors: Chen Fang, Jimei Yang, Zhaowen Wang, Ming-hsuan Yang
    Abstract:

    Recent progresses on deep discriminative and generative modeling have shown promising results on texture synthesis. However, existing feed-Forward based methods trade off generality for efficiency, which suffer from many issues, such as shortage of generality (i.e., build one network per texture), lack of diversity (i.e., always produce visually identical output) and suboptimality (i.e., generate less satisfying visual effects). In this work, we focus on solving these issues for improved texture synthesis. We propose a deep generative feed-Forward network which enables efficient synthesis of multiple textures within one single network and meaningful interpolation between them. Meanwhile, a suite of important techniques are introduced to achieve better convergence and diversity. With extensive experiments, we demonstrate the effectiveness of the proposed model and techniques for synthesizing a large number of textures and show its applications with the stylization.

Chen Fang - One of the best experts on this subject based on the ideXlab platform.

  • Diversified Texture Synthesis with Feed-Forward Networks
    arXiv: Computer Vision and Pattern Recognition, 2017
    Co-Authors: Chen Fang, Jimei Yang, Zhaowen Wang, Ming-hsuan Yang
    Abstract:

    Recent progresses on deep discriminative and generative modeling have shown promising results on texture synthesis. However, existing feed-Forward based methods trade off generality for efficiency, which suffer from many issues, such as shortage of generality (i.e., build one network per texture), lack of diversity (i.e., always produce visually identical output) and suboptimality (i.e., generate less satisfying visual effects). In this work, we focus on solving these issues for improved texture synthesis. We propose a deep generative feed-Forward network which enables efficient synthesis of multiple textures within one single network and meaningful interpolation between them. Meanwhile, a suite of important techniques are introduced to achieve better convergence and diversity. With extensive experiments, we demonstrate the effectiveness of the proposed model and techniques for synthesizing a large number of textures and show its applications with the stylization.

  • CVPR - Diversified Texture Synthesis with Feed-Forward Networks
    2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017
    Co-Authors: Chen Fang, Jimei Yang, Zhaowen Wang, Ming-hsuan Yang
    Abstract:

    Recent progresses on deep discriminative and generative modeling have shown promising results on texture synthesis. However, existing feed-Forward based methods trade off generality for efficiency, which suffer from many issues, such as shortage of generality (i.e., build one network per texture), lack of diversity (i.e., always produce visually identical output) and suboptimality (i.e., generate less satisfying visual effects). In this work, we focus on solving these issues for improved texture synthesis. We propose a deep generative feed-Forward network which enables efficient synthesis of multiple textures within one single network and meaningful interpolation between them. Meanwhile, a suite of important techniques are introduced to achieve better convergence and diversity. With extensive experiments, we demonstrate the effectiveness of the proposed model and techniques for synthesizing a large number of textures and show its applications with the stylization.

Eric Thierry - One of the best experts on this subject based on the ideXlab platform.

  • Tight performance bounds in the worst-case analysis of feed-Forward Networks *
    Discrete Event Dynamic Systems, 2016
    Co-Authors: Anne Bouillard, Eric Thierry
    Abstract:

    Network Calculus theory aims at evaluating worst-case performances in communication Networks. It provides methods to analyze models where the traffic and the services are constrained by some minimum and/or maximum envelopes (arrival/service curves). While new applications come Forward , a challenging and inescapable issue remains open: achieving tight analyzes of Networks with aggregate multiplexing. The theory offers efficient methods to bound maximum end-to-end delays or local backlogs. However as shown in a recent breakthrough paper [28], those bounds can be arbitrarily far from the exact worst-case values, even in seemingly simple feed-Forward Networks (two flows and two servers), under blind multiplexing (i.e. no information about the scheduling policies, except FIFO per flow). For now, only a network with three flows and three servers, as well as a tandem network called sink tree, have been analyzed tightly. We describe the first algorithm which computes the maximum end-to-end delay for a given flow, as well as the maximum backlog at a server, for any feed-Forward network under blind multiplexing, with piecewise affine concave arrival curves and piecewise affine convex service curves. Its computational complexity may look expensive (possibly super-exponential), but we show that the problem is intrinsically difficult (NP-hard). Fortunately we show that in some cases, like tandem Networks with cross-traffic interfering along intervals of servers, the complexity becomes polynomial. We also compare ourselves to the previous approaches and discuss the problems left open.

  • Tight performance bounds in the worst-case analysis of feed-Forward Networks
    2010
    Co-Authors: Anne Bouillard, Laurent Jouhet, Eric Thierry
    Abstract:

    Network Calculus theory aims at evaluating worst-case performances in communication Networks. It provides methods to analyze models where the traffic and the services are constrained by some minimum and/or maximum envelopes (service/arrival curves). While new applications come Forward, a challenging and inescapable issue remains open: achieving tight analyzes of Networks with aggregate multiplexing. The theory offers efficient methods to bound maximum end-to-end delays or local backlogs. However as shown recently, those bounds can be arbitrarily far from the exact worst-case values, even in seemingly simple feed-Forward Networks (two flows and two servers), under blind multiplexing (i.e. no information about the scheduling policies, except FIFO per flow). For now, only a network with three flows and three servers, as well as a tandem network called sink tree, have been analyzed tightly. We describe the first algorithm which computes the maximum end-to-end delay for a given flow, as well as the maximum backlog at a server, for any feed-Forward network under blind multiplex-ing, with concave arrival curves and convex service curves. Its computational complexity may look expensive (possibly super-exponential), but we show that the problem is intrinsically difficult (NP-hard). Fortunately we show that in some cases, like tandem Networks with cross-traffic interfering along intervals of servers, the complexity becomes polynomial. We also compare ourselves to the previous approaches and discuss the problems left open.

  • INFOCOM - Tight Performance Bounds in the Worst-Case Analysis of Feed-Forward Networks
    2010 Proceedings IEEE INFOCOM, 2010
    Co-Authors: Anne Bouillard, Laurent Jouhet, Eric Thierry
    Abstract:

    Network Calculus theory aims at evaluating worst-case performances in communication Networks. It provides methods to analyze models where the traffic and the services are constrained by some minimum and/or maximum envelopes (service/arrival curves). While new applications come Forward, a challenging and inescapable issue remains open: achieving tight analyzes of Networks with aggregate multiplexing. The theory offers efficient methods to bound maximum end-to-end delays or local backlogs. However as shown recently, those bounds can be arbitrarily far from the exact worst-case values, even in seemingly simple feed-Forward Networks (two flows and two servers), under blind multiplexing (i.e. no information about the scheduling policies, except FIFO per flow). For now, only a network with three flows and three servers, as well as a tandem network called sink tree, have been analyzed tightly. We describe the first algorithm which computes the maximum end-to-end delay for a given flow, as well as the maximum backlog at a server, for any feed-Forward network under blind multiplexing, with concave arrival curves and convex service curves. Its computational complexity may look expensive (possibly super-exponential), but we show that the problem is intrinsically difficult (NP-hard). Fortunately we show that in some cases, like tandem Networks with cross-traffic interfering along intervals of servers, the complexity becomes polynomial. We also compare ourselves to the previous approaches and discuss the problems left open.

  • Tight performance bounds in the worst-case analysis of feed-Forward Networks
    2009
    Co-Authors: Anne Bouillard, Laurent Jouhet, Eric Thierry
    Abstract:

    Network Calculus theory aims at evaluating worst-case performances in communication Networks. It provides methods to analyze models where the traffic and the services are constrained by some minimum and/or maximum envelopes (arrival/service curves). While new applications come Forward, a challenging and inescapable issue remains open: achieving tight analyzes of Networks with aggregate multiplexing. The theory offers efficient methods to bound maximum end-to-end delays or local backlogs. However as shown in a recent breakthrough paper [Schmitt-Infocom08], those bounds can be arbitrarily far from the exact worst-case values, even in seemingly simple feed-Forward Networks (two flows and two servers), under blind multiplexing (i.e. no information about the scheduling policies, except FIFO per flow). For now, only a network with three flows and three servers, as well as a tandem network called sink tree, have been analyzed tightly. We describe the first algorithm which computes the maximum end-to-end delay for a given flow, as well as the maximum backlog at a server, for {\em any} feed-Forward network under blind multiplexing, with concave arrival curves and convex service curves. Its computational complexity may look expensive (possibly super-exponential), but we show that the problem is intrinsically difficult (NP-hard). Fortunately we show that in some cases, like tandem Networks with cross-traffic interfering along intervals of servers, the complexity becomes polynomial. We also compare ourselves to the previous approaches and discuss the problems left open.

Sharad Singhal - One of the best experts on this subject based on the ideXlab platform.

  • ICASSP - Training feed-Forward Networks with the extended Kalman algorithm
    International Conference on Acoustics Speech and Signal Processing, 1
    Co-Authors: Sharad Singhal
    Abstract:

    It is shown that training feed-Forward nets can be viewed as a system identification problem for a nonlinear dynamic system. For linear dynamic systems, the Kalman filter is known to produce an optimal estimator. Extended versions of the Kalman algorithm can be used to train feed-Forward Networks. The performance of the Kalman algorithm is examined using artificially constructed examples with two inputs, and it is found that the algorithm typically converges in a few iterations. Backpropagation is used on the same examples, and the Kalman algorithm invariably converges in fewer iterations. For the XOR problem, backpropagation fails to converge on any of the cases considered, whereas the Kalman algorithm is able to find solutions with the same network configurations. >

Markus Diesmann - One of the best experts on this subject based on the ideXlab platform.

  • bifurcation analysis of synchronization dynamics in cortical feed Forward Networks in novel coordinates
    BMC Neuroscience, 2009
    Co-Authors: Tilo Schwalger, Sven Goedeke, Markus Diesmann
    Abstract:

    In a synfire chain [1], synchronous activity in one group of neurons can excite neurons of the next group to fire synchronously themselves. If this mechanism repeats itself from group to group, a "pulse packet" of spiking activity can travel down the chain. For a homogeneous chain, the spike packet profile of one group is uniquely mapped to the packet profile of the successive group, thereby establishing a map for the packet dynamics in the space of pulse-shaped functions. A stable packet corresponds to a stable fixed point of this infinite-dimensional map.

  • Consequences of realistic network size on the stability of embedded synfire chains
    Neurocomputing, 2004
    Co-Authors: Tom Tetzlaff, Abigail Morrison, Theo Geisel, Markus Diesmann
    Abstract:

    Abstract Cortical activity in vivo is characterized by asynchronous irregular spiking. Additionally, precise spike synchronization is observed with respect to the experimental protocol. Attempting to model this behavior, theoretical studies have focused on two extreme cases: random and feed-Forward Networks (synfire chains). Here, we combine both descriptions by successively converting an isolated synfire chain into a completely embedded one. This method systematically reveals the effects of different aspects of the embedding scheme on the stability of the system. At realistic network sizes common-input correlations play a major role. Surprisingly, their impact is reduced by the dynamics of the embedding recurrent network.

  • The ground state of cortical feed-Forward Networks
    Neurocomputing, 2002
    Co-Authors: Tom Tetzlaff, Theo Geisel, Markus Diesmann
    Abstract:

    Abstract The occurrence of spatio-temporal spike patterns in the cortex is explained by models of divergent/convergent feed-Forward subNetworks—synfire chains. Their excited mode is characterized by spike volleys propagating from one neuron group to the next. We demonstrate the existence of an upper bound for group size: above a critical value synchronous activity develops spontaneously from random fluctuations. Stability of the ground state, in which neurons independently fire at low rates, is lost. Comparison of an analytic rate model with network simulations shows that the transition from the asynchronous into the synchronous regime is driven by an instability in rate dynamics.