We consider the problem of decontaminating galaxy spectra in the context of the EUCLID space mission. The spectra of neighboring astronomical objects being spatially mixed, a source separation method should be used to...
详细信息
ISBN:
(纸本)9781479981311
We consider the problem of decontaminating galaxy spectra in the context of the EUCLID space mission. The spectra of neighboring astronomical objects being spatially mixed, a source separation method should be used to separate them. Here, we propose a new method based on the fusion of information between first and second-order spectra generated by a grism. Using the optical properties, we propose a regularized criterion and a gradient algorithm to optimize it. The tests using noisy realistic simulated data show that our method leads to better results than a method only based on second-order information.
We propose a distributed differentially-private canonical correlation analysis (CCA) algorithm to use on multi-view data. CCA finds a subspace for each view such that projecting the views onto these subspaces simultan...
详细信息
ISBN:
(纸本)9781479981311
We propose a distributed differentially-private canonical correlation analysis (CCA) algorithm to use on multi-view data. CCA finds a subspace for each view such that projecting the views onto these subspaces simultaneously reduces the dimension and maximizes correlation. In applications involving privacy-sensitive data, such as medical imaging, distributed privacy-preserving algorithms can let data holders maintain local control of their data while participating in joint computations with other data holders. Differential privacy is a framework for quantifying the privacy risk in such settings. However, conventional distributed differentially-private algorithms introduce more noise to guarantee a given level of privacy compared to their centralized counterparts. Our differentially-private CCA employs a noise-reduction strategy to achieve the same utility level as CCA on centralized data. Experiments on synthetic and real data show the benefit of our approach over conventional methods.
In this work, we investigate residual neural network representations for the identification and forecasting of dynamical systems. We propose a novel architecture that jointly learns the dynamical model and the associa...
详细信息
ISBN:
(纸本)9781479981311
In this work, we investigate residual neural network representations for the identification and forecasting of dynamical systems. We propose a novel architecture that jointly learns the dynamical model and the associated Runge-Kutta integration scheme. We demonstrate the relevance of the proposed architecture with respect to learning-based state-of-the-art approaches in the identification and forecasting of chaotic dynamics when provided with training data with low temporal sampling rates.
The digital pen is the main tool for obtaining online signature writing information. Since the measuring coordinate system of the digital pen is established at the pen body, it must be held at the same position each t...
详细信息
ISBN:
(纸本)9781728136608
The digital pen is the main tool for obtaining online signature writing information. Since the measuring coordinate system of the digital pen is established at the pen body, it must be held at the same position each time to ensure the consistency of the data. In this paper, by combining the tilt sensor with the force, a pen-based solution using high-quality sensor is designed, and the accuracy of the data is verified. Finally, the limitation of the digital pen during writing is eliminated.
data augmentation is crucial to improving the performance of deep neural networks by helping the model avoid overfitting and improve its generalization. In automatic speech recognition, previous work proposed several ...
详细信息
ISBN:
(纸本)9781479981311
data augmentation is crucial to improving the performance of deep neural networks by helping the model avoid overfitting and improve its generalization. In automatic speech recognition, previous work proposed several approaches to augment data by performing speed perturbation or spectral transformation. Since data augmented in this manner has similar acoustic representations as the original data, it has limited advantage in improving generalization of the acoustic model. In order to avoid generating data with limited diversity, we propose a voice conversion approach using a generative model (WaveNet), which generates a new utterance by transforming an utterance to a given target voice. Our method synthesizes speech with diverse pitch patterns by minimizing the use of acoustic features. With the Wall Street Journal dataset, we verify that our method led to better generalization compared to other data augmentation techniques such as speed perturbation and WORLD-based voice conversion. In addition, when combined with the speed perturbation technique, the two methods complement each other to further improve performance of the acoustic model.
Due to the limited radius of the arc, traditional ground-based Arc SAR often suffers from poor azimuth resolution. To provide high-resolution wide-swath imaging, a circular-track ringmap SAR equipped on multi-rotors U...
详细信息
ISBN:
(纸本)9781728136608
Due to the limited radius of the arc, traditional ground-based Arc SAR often suffers from poor azimuth resolution. To provide high-resolution wide-swath imaging, a circular-track ringmap SAR equipped on multi-rotors UAV platform is proposed in this paper. In this paper, we put forward "ring-map mode" for the first time to describe a new data collection geometry. We begin with a brief description of the geometry model and discussion of its achievable image resolution. Then, an image formation processing flowchart based on sub-aperture processing strategy is proposed. Finally, simulation dataprocessing verifies the effectiveness of the proposed approach.
Electrocardiogram (ECG) is the recording of the heart electrical activity and used to diagnose heart disease nowadays. The diagnosis requires a large amount of time for acquiring enough multi-channel data normally. Th...
详细信息
ISBN:
(纸本)9781479981311
Electrocardiogram (ECG) is the recording of the heart electrical activity and used to diagnose heart disease nowadays. The diagnosis requires a large amount of time for acquiring enough multi-channel data normally. Thus storage and transmission of 12 lead ECG data will result in massive cost. In this work, we propose a multi-channel ECG lossless compression which uses the adaptive linear prediction for intra and inter channel decorrelation. The proposed technique is based on the adaptive Golomb-Rice codec for entropy coding with adaptive linear prediction. Thus the coefficient of linear prediction and Golomb-Rice codec will make self-adjustment during the process. Finally we evaluate the proposed algorithm with MIT-BIH Arrhythmia database for single-channel compression, and PTB database for multi-channel compression.
Based on recent random matrix advances in the analysis of kernel methods for classification and clustering, this paper proposes the study of large kernel methods for a wide class of random inputs, i.e., concentrated d...
详细信息
ISBN:
(纸本)9781479981311
Based on recent random matrix advances in the analysis of kernel methods for classification and clustering, this paper proposes the study of large kernel methods for a wide class of random inputs, i.e., concentrated data, which are more generic than Gaussian mixtures. The concentration assumption is motivated by the fact that one can use generative models to design complex data structures, through Lipschitzally transformed concentrated vectors (e.g., Gaussian) which remain concentrated vectors. Applied to spectral clustering, we demonstrate that our theoretical findings closely match the behavior of large kernel matrices, when considering the fedin data as CNN representations of GAN-generated images (i.e., concentrated vectors by design).
Neural dialog models often lack robustness to anomalous user input and produce inappropriate responses which leads to frustrating user experience. Although there are a set of prior approaches to out-of-domain (OOD) ut...
详细信息
ISBN:
(纸本)9781479981311
Neural dialog models often lack robustness to anomalous user input and produce inappropriate responses which leads to frustrating user experience. Although there are a set of prior approaches to out-of-domain (OOD) utterance detection, they share a few restrictions: they rely on OOD data or multiple sub-domains, and their OOD detection is context-independent which leads to suboptimal performance in a dialog. The goal of this paper is to propose a novel OOD detection method that does not require OOD data by utilizing counterfeit OOD turns in the context of a dialog. For the sake of fostering further research, we also release new dialog datasets which are 3 publicly available dialog corpora augmented with OOD turns in a controllable way. Our method outperforms state-of-the-art dialog models equipped with a conventional OOD detection mechanism by a large margin in the presence of OOD utterances.
Multimodal data fusion is an important aspect of many object localization and tracking frameworks that rely on sensory observations from different sources. A prominent example is audiovisual speaker localization, wher...
详细信息
ISBN:
(纸本)9781479981311
Multimodal data fusion is an important aspect of many object localization and tracking frameworks that rely on sensory observations from different sources. A prominent example is audiovisual speaker localization, where the incorporation of visual information has shown to benefit overall performance, especially in adverse acoustic conditions. Recently, the notion of dynamic stream weights as an efficient data fusion technique has been introduced into this field. Originally proposed in the context of audiovisual automatic speech recognition, dynamic stream weights allow for effective sensory-level data fusion on a per-frame basis, if reliability measures for the individual sensory streams are available. This study proposes a learning framework for dynamic stream weights based on natural evolution strategies, which does not require the explicit computation of oracle information. An experimental evaluation based on recorded audiovisual sequences shows that the proposed approach outperforms conventional methods based on supervised training in terms of localization performance.
暂无评论