The Experts below are selected from a list of 144 Experts worldwide ranked by ideXlab platform

Heng Huang - One of the best experts on this subject based on the ideXlab platform.

  • deep clustering via joint convolutional autoencoder Embedding and relative entropy minimization
    International Conference on Computer Vision, 2017
    Co-Authors: Kamran Ghasedi Dizaji, Amirhossein Herandi, Cheng Deng, Heng Huang
    Abstract:

    In this paper, we propose a new clustering model, called DEeP Embedded Regularized ClusTering (DEPICT), which efficiently maps data into a discriminative Embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of a multinomial logistic regression Function stacked on top of a multi-layer convolutional autoencoder. We define a clustering objective Function using relative entropy (KL divergence) minimization, regularized by a prior for the frequency of cluster assignments. An alternating strategy is then derived to optimize the objective by updating parameters and estimating cluster assignments. Furthermore, we employ the reconstruction loss Functions in our autoencoder, as a data-dependent regularization term, to prevent the deep Embedding Function from overfitting. In order to benefit from end-to-end optimization and eliminate the necessity for layer-wise pre-training, we introduce a joint learning framework to minimize the unified clustering and reconstruction loss Functions together and train all network layers simultaneously. Experimental results indicate the superiority and faster running time of DEPICT in real-world clustering tasks, where no labeled data is available for hyper-parameter tuning.

  • deep clustering via joint convolutional autoencoder Embedding and relative entropy minimization
    arXiv: Learning, 2017
    Co-Authors: Kamran Ghasedi Dizaji, Amirhossein Herandi, Cheng Deng, Heng Huang
    Abstract:

    Image clustering is one of the most important computer vision applications, which has been extensively studied in literature. However, current clustering methods mostly suffer from lack of efficiency and scalability when dealing with large-scale and high-dimensional data. In this paper, we propose a new clustering model, called DEeP Embedded RegularIzed ClusTering (DEPICT), which efficiently maps data into a discriminative Embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of a multinomial logistic regression Function stacked on top of a multi-layer convolutional autoencoder. We define a clustering objective Function using relative entropy (KL divergence) minimization, regularized by a prior for the frequency of cluster assignments. An alternating strategy is then derived to optimize the objective by updating parameters and estimating cluster assignments. Furthermore, we employ the reconstruction loss Functions in our autoencoder, as a data-dependent regularization term, to prevent the deep Embedding Function from overfitting. In order to benefit from end-to-end optimization and eliminate the necessity for layer-wise pretraining, we introduce a joint learning framework to minimize the unified clustering and reconstruction loss Functions together and train all network layers simultaneously. Experimental results indicate the superiority and faster running time of DEPICT in real-world clustering tasks, where no labeled data is available for hyper-parameter tuning.

Kamran Ghasedi Dizaji - One of the best experts on this subject based on the ideXlab platform.

  • deep clustering via joint convolutional autoencoder Embedding and relative entropy minimization
    International Conference on Computer Vision, 2017
    Co-Authors: Kamran Ghasedi Dizaji, Amirhossein Herandi, Cheng Deng, Heng Huang
    Abstract:

    In this paper, we propose a new clustering model, called DEeP Embedded Regularized ClusTering (DEPICT), which efficiently maps data into a discriminative Embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of a multinomial logistic regression Function stacked on top of a multi-layer convolutional autoencoder. We define a clustering objective Function using relative entropy (KL divergence) minimization, regularized by a prior for the frequency of cluster assignments. An alternating strategy is then derived to optimize the objective by updating parameters and estimating cluster assignments. Furthermore, we employ the reconstruction loss Functions in our autoencoder, as a data-dependent regularization term, to prevent the deep Embedding Function from overfitting. In order to benefit from end-to-end optimization and eliminate the necessity for layer-wise pre-training, we introduce a joint learning framework to minimize the unified clustering and reconstruction loss Functions together and train all network layers simultaneously. Experimental results indicate the superiority and faster running time of DEPICT in real-world clustering tasks, where no labeled data is available for hyper-parameter tuning.

  • deep clustering via joint convolutional autoencoder Embedding and relative entropy minimization
    arXiv: Learning, 2017
    Co-Authors: Kamran Ghasedi Dizaji, Amirhossein Herandi, Cheng Deng, Heng Huang
    Abstract:

    Image clustering is one of the most important computer vision applications, which has been extensively studied in literature. However, current clustering methods mostly suffer from lack of efficiency and scalability when dealing with large-scale and high-dimensional data. In this paper, we propose a new clustering model, called DEeP Embedded RegularIzed ClusTering (DEPICT), which efficiently maps data into a discriminative Embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of a multinomial logistic regression Function stacked on top of a multi-layer convolutional autoencoder. We define a clustering objective Function using relative entropy (KL divergence) minimization, regularized by a prior for the frequency of cluster assignments. An alternating strategy is then derived to optimize the objective by updating parameters and estimating cluster assignments. Furthermore, we employ the reconstruction loss Functions in our autoencoder, as a data-dependent regularization term, to prevent the deep Embedding Function from overfitting. In order to benefit from end-to-end optimization and eliminate the necessity for layer-wise pretraining, we introduce a joint learning framework to minimize the unified clustering and reconstruction loss Functions together and train all network layers simultaneously. Experimental results indicate the superiority and faster running time of DEPICT in real-world clustering tasks, where no labeled data is available for hyper-parameter tuning.

Efthimios Kaxiras - One of the best experts on this subject based on the ideXlab platform.

  • embedded atom method potentials employing a faithful density representation
    Modelling and Simulation in Materials Science and Engineering, 2007
    Co-Authors: Pavlin D Mitev, G A Evangelakis, Efthimios Kaxiras
    Abstract:

    We present an approach for deriving embedded atom method energy Functionals which employs a faithful representation of the valence electron that reproduces ab initio electronic structure calculations. This approach offers the possibility of improved accuracy and versatility over existing methods. Moreover, the approach has a distinct advantage for coupling to more accurate methods in the context of multiscale schemes. The Embedding Function is based on first breaking down the electronic density to individual atomic contributions and then designing an interatomic Function which captures the interaction between the atomic contributions towards formation of the interatomic bonds. We use Al as a prototypical metallic solid to illustrate the application of the method and we employ density Functional theory (DFT) to calculate the electronic charge densities and energies for determining the values of fitting parameters. We validate the approach by reproducing adequately experimental data for the cohesive energy, bulk modulus, elastic constants and dynamical properties at finite temperatures, obtained by molecular dynamics simulations.

Tadao Onzawa - One of the best experts on this subject based on the ideXlab platform.

  • development of modified embedded atom method for a bcc metal lithium
    Modelling and Simulation in Materials Science and Engineering, 2003
    Co-Authors: Xiaoying Yuan, Kunio Takahashi, Yingchen Yin, Tadao Onzawa
    Abstract:

    A new scheme of modified embedded atom method (MEAM) is proposed in this paper. The analytic form of the Embedding Function is modified. All the parameters of MEAM have been reset by relating them with bulk properties and some non-bulk properties, for example, the bond length of a dimer and the change of surface interlayer distance. The new scheme has been applied to calculate the elastic stiffness of crystal, the vacancy formation energy and some properties of non-bulk systems such as the surface energies for low index crystal faces, the bond length and the binding energy for a dimer. The results are compared with the experimental data and get a fairly good agreement.

Cheng Deng - One of the best experts on this subject based on the ideXlab platform.

  • deep clustering via joint convolutional autoencoder Embedding and relative entropy minimization
    International Conference on Computer Vision, 2017
    Co-Authors: Kamran Ghasedi Dizaji, Amirhossein Herandi, Cheng Deng, Heng Huang
    Abstract:

    In this paper, we propose a new clustering model, called DEeP Embedded Regularized ClusTering (DEPICT), which efficiently maps data into a discriminative Embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of a multinomial logistic regression Function stacked on top of a multi-layer convolutional autoencoder. We define a clustering objective Function using relative entropy (KL divergence) minimization, regularized by a prior for the frequency of cluster assignments. An alternating strategy is then derived to optimize the objective by updating parameters and estimating cluster assignments. Furthermore, we employ the reconstruction loss Functions in our autoencoder, as a data-dependent regularization term, to prevent the deep Embedding Function from overfitting. In order to benefit from end-to-end optimization and eliminate the necessity for layer-wise pre-training, we introduce a joint learning framework to minimize the unified clustering and reconstruction loss Functions together and train all network layers simultaneously. Experimental results indicate the superiority and faster running time of DEPICT in real-world clustering tasks, where no labeled data is available for hyper-parameter tuning.

  • deep clustering via joint convolutional autoencoder Embedding and relative entropy minimization
    arXiv: Learning, 2017
    Co-Authors: Kamran Ghasedi Dizaji, Amirhossein Herandi, Cheng Deng, Heng Huang
    Abstract:

    Image clustering is one of the most important computer vision applications, which has been extensively studied in literature. However, current clustering methods mostly suffer from lack of efficiency and scalability when dealing with large-scale and high-dimensional data. In this paper, we propose a new clustering model, called DEeP Embedded RegularIzed ClusTering (DEPICT), which efficiently maps data into a discriminative Embedding subspace and precisely predicts cluster assignments. DEPICT generally consists of a multinomial logistic regression Function stacked on top of a multi-layer convolutional autoencoder. We define a clustering objective Function using relative entropy (KL divergence) minimization, regularized by a prior for the frequency of cluster assignments. An alternating strategy is then derived to optimize the objective by updating parameters and estimating cluster assignments. Furthermore, we employ the reconstruction loss Functions in our autoencoder, as a data-dependent regularization term, to prevent the deep Embedding Function from overfitting. In order to benefit from end-to-end optimization and eliminate the necessity for layer-wise pretraining, we introduce a joint learning framework to minimize the unified clustering and reconstruction loss Functions together and train all network layers simultaneously. Experimental results indicate the superiority and faster running time of DEPICT in real-world clustering tasks, where no labeled data is available for hyper-parameter tuning.