The Experts below are selected from a list of 270 Experts worldwide ranked by ideXlab platform

Linlin Chao - One of the best experts on this subject based on the ideXlab platform.

  • long short term memory recurrent neural network based multimodal dimensional emotion recognition
    ACM Multimedia, 2015
    Co-Authors: Linlin Chao, Minghao Yang, Ya Li
    Abstract:

    This paper presents our effort to the Audio/Visual+ Emotion Challenge (AV+EC2015), whose goal is to predict the continuous values of the emotion dimensions arousal and valence from audio, visual and physiology modalities. The state of art classifier for dimensional recognition, long short term memory recurrent neural network (LSTM-RNN) is utilized. Except regular LSTM-RNN Prediction architecture, two techniques are investigated for dimensional emotion recognition problem. The first one is e -insensitive loss is utilized as the loss function to optimize. Compared to squared loss function, which is the most widely used loss function for dimension emotion recognition, e -insensitive loss is more robust for the label noises and it can ignore small errors to get stronger correlation between Predictions and labels. The other one is temporal pooling. This technique enables temporal modeling in the input features and increases the diversity of the features fed into the Forward Prediction architecture. Experiments results show the efficiency of key points of the proposed method and competitive results are obtained.

  • long short term memory recurrent neural network based multimodal dimensional emotion recognition
    ACM Multimedia, 2015
    Co-Authors: Linlin Chao, Minghao Yang, Jianhua Tao, Zhengqi Wen
    Abstract:

    This paper presents our effort to the Audio/Visual+ Emotion Challenge (AV+EC2015), whose goal is to predict the continuous values of the emotion dimensions arousal and valence from audio, visual and physiology modalities. The state of art classifier for dimensional recognition, long short term memory recurrent neural network (LSTM-RNN) is utilized. Except regular LSTM-RNN Prediction architecture, two techniques are investigated for dimensional emotion recognition problem. The first one is e -insensitive loss is utilized as the loss function to optimize. Compared to squared loss function, which is the most widely used loss function for dimension emotion recognition, e -insensitive loss is more robust for the label noises and it can ignore small errors to get stronger correlation between Predictions and labels. The other one is temporal pooling. This technique enables temporal modeling in the input features and increases the diversity of the features fed into the Forward Prediction architecture. Experiments results show the efficiency of key points of the proposed method and competitive results are obtained.

Ya Li - One of the best experts on this subject based on the ideXlab platform.

  • long short term memory recurrent neural network based multimodal dimensional emotion recognition
    ACM Multimedia, 2015
    Co-Authors: Linlin Chao, Minghao Yang, Ya Li
    Abstract:

    This paper presents our effort to the Audio/Visual+ Emotion Challenge (AV+EC2015), whose goal is to predict the continuous values of the emotion dimensions arousal and valence from audio, visual and physiology modalities. The state of art classifier for dimensional recognition, long short term memory recurrent neural network (LSTM-RNN) is utilized. Except regular LSTM-RNN Prediction architecture, two techniques are investigated for dimensional emotion recognition problem. The first one is e -insensitive loss is utilized as the loss function to optimize. Compared to squared loss function, which is the most widely used loss function for dimension emotion recognition, e -insensitive loss is more robust for the label noises and it can ignore small errors to get stronger correlation between Predictions and labels. The other one is temporal pooling. This technique enables temporal modeling in the input features and increases the diversity of the features fed into the Forward Prediction architecture. Experiments results show the efficiency of key points of the proposed method and competitive results are obtained.

Zhengqi Wen - One of the best experts on this subject based on the ideXlab platform.

  • long short term memory recurrent neural network based multimodal dimensional emotion recognition
    ACM Multimedia, 2015
    Co-Authors: Linlin Chao, Minghao Yang, Jianhua Tao, Zhengqi Wen
    Abstract:

    This paper presents our effort to the Audio/Visual+ Emotion Challenge (AV+EC2015), whose goal is to predict the continuous values of the emotion dimensions arousal and valence from audio, visual and physiology modalities. The state of art classifier for dimensional recognition, long short term memory recurrent neural network (LSTM-RNN) is utilized. Except regular LSTM-RNN Prediction architecture, two techniques are investigated for dimensional emotion recognition problem. The first one is e -insensitive loss is utilized as the loss function to optimize. Compared to squared loss function, which is the most widely used loss function for dimension emotion recognition, e -insensitive loss is more robust for the label noises and it can ignore small errors to get stronger correlation between Predictions and labels. The other one is temporal pooling. This technique enables temporal modeling in the input features and increases the diversity of the features fed into the Forward Prediction architecture. Experiments results show the efficiency of key points of the proposed method and competitive results are obtained.

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

  • long short term memory recurrent neural network based multimodal dimensional emotion recognition
    ACM Multimedia, 2015
    Co-Authors: Linlin Chao, Minghao Yang, Ya Li
    Abstract:

    This paper presents our effort to the Audio/Visual+ Emotion Challenge (AV+EC2015), whose goal is to predict the continuous values of the emotion dimensions arousal and valence from audio, visual and physiology modalities. The state of art classifier for dimensional recognition, long short term memory recurrent neural network (LSTM-RNN) is utilized. Except regular LSTM-RNN Prediction architecture, two techniques are investigated for dimensional emotion recognition problem. The first one is e -insensitive loss is utilized as the loss function to optimize. Compared to squared loss function, which is the most widely used loss function for dimension emotion recognition, e -insensitive loss is more robust for the label noises and it can ignore small errors to get stronger correlation between Predictions and labels. The other one is temporal pooling. This technique enables temporal modeling in the input features and increases the diversity of the features fed into the Forward Prediction architecture. Experiments results show the efficiency of key points of the proposed method and competitive results are obtained.

  • long short term memory recurrent neural network based multimodal dimensional emotion recognition
    ACM Multimedia, 2015
    Co-Authors: Linlin Chao, Minghao Yang, Jianhua Tao, Zhengqi Wen
    Abstract:

    This paper presents our effort to the Audio/Visual+ Emotion Challenge (AV+EC2015), whose goal is to predict the continuous values of the emotion dimensions arousal and valence from audio, visual and physiology modalities. The state of art classifier for dimensional recognition, long short term memory recurrent neural network (LSTM-RNN) is utilized. Except regular LSTM-RNN Prediction architecture, two techniques are investigated for dimensional emotion recognition problem. The first one is e -insensitive loss is utilized as the loss function to optimize. Compared to squared loss function, which is the most widely used loss function for dimension emotion recognition, e -insensitive loss is more robust for the label noises and it can ignore small errors to get stronger correlation between Predictions and labels. The other one is temporal pooling. This technique enables temporal modeling in the input features and increases the diversity of the features fed into the Forward Prediction architecture. Experiments results show the efficiency of key points of the proposed method and competitive results are obtained.

Susan Holmes - One of the best experts on this subject based on the ideXlab platform.

  • exact sequence variants should replace operational taxonomic units in marker gene data analysis
    The ISME Journal, 2017
    Co-Authors: Benjamin J Callahan, Paul J Mcmurdie, Susan Holmes
    Abstract:

    Recent advances have made it possible to analyze high-throughput marker-gene sequencing data without resorting to the customary construction of molecular operational taxonomic units (OTUs): clusters of sequencing reads that differ by less than a fixed dissimilarity threshold. New methods control errors sufficiently such that amplicon sequence variants (ASVs) can be resolved exactly, down to the level of single-nucleotide differences over the sequenced gene region. The benefits of finer resolution are immediately apparent, and arguments for ASV methods have focused on their improved resolution. Less obvious, but we believe more important, are the broad benefits that derive from the status of ASVs as consistent labels with intrinsic biological meaning identified independently from a reference database. Here we discuss how these features grant ASVs the combined advantages of closed-reference OTUs—including computational costs that scale linearly with study size, simple merging between independently processed data sets, and Forward Prediction—and of de novo OTUs—including accurate measurement of diversity and applicability to communities lacking deep coverage in reference databases. We argue that the improvements in reusability, reproducibility and comprehensiveness are sufficiently great that ASVs should replace OTUs as the standard unit of marker-gene analysis and reporting.

  • exact sequence variants should replace operational taxonomic units in marker gene data analysis
    bioRxiv, 2017
    Co-Authors: Benjamin J Callahan, Paul J Mcmurdie, Susan Holmes
    Abstract:

    Recent advances have made it possible to analyze high-throughput marker-gene sequencing data without resorting to the customary construction of molecular operational taxonomic units (OTUs): clusters of sequencing reads that differ by less than a fixed dissimilarity threshold. New methods control errors sufficiently that sequence variants (SVs) can be resolved exactly, down to the level of single-nucleotide differences over the sequenced gene region. The benefits of finer taxonomic resolution are immediately apparent, and arguments for SV methods have focused on their improved resolution. Less obvious, but we believe more important, are the broad benefits deriving from the status of SVs as consistent labels with intrinsic biological meaning identified independently from a reference database . Here we discuss how those features grant SVs the combined advantages of closed-reference OTUs -- including computational costs that scale linearly with study size, simple merging between independently processed datasets, and Forward Prediction -- and of de novo OTUs -- including accurate diversity measurement and applicability to communities lacking deep coverage in reference databases. We argue that the improvements in reusability, reproducibility and comprehensiveness are sufficiently great that SVs should replace OTUs as the standard unit of marker gene analysis and reporting.