arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.06704cs.LGstat.ML
收藏

带噪声发生器的端到端Sinkhorn自编码器

End-to-end Sinkhorn Autoencoder with Noise Generator

Warsaw University of Technology(华沙理工大学);CERN(欧洲核子研究组织);Tooploox

Kamil Deja, Jan Dubiński, Piotr Nowak, Sandro Wenzel, Tomasz Trzciński

摘要:

在这项工作中,我们提出了一种新的端到端的sinkhorn自动编码器与噪声发生器有效的数据汇集模拟。模拟旨在收集实验数据的过程对于多种现实应用至关重要,包括核医学,天文学和高能物理。当代的方法,如蒙特卡罗算法,提供高保真度的结果,在高计算成本的代价。采取多种尝试来减少这种负担,例如,使用基于生成对抗网络或变分自编码器的生成方法。虽然这些方法要快得多,但它们在训练中通常不稳定,并且不允许从整个数据分布中采样。为了解决这些缺点,我们引入了一种称为端到端Sinkhorn Autoencoder的新方法,该方法利用Sinkhorn算法来显式对齐编码的真实数据示例和生成的噪声的分布。更准确地说,我们通过添加一个确定性的神经网络训练,以映射噪声从一个已知的分布到自动编码器的潜在空间表示数据分布,扩展自动编码器的架构。我们共同优化整个模型。我们的方法优于竞争的方法在一个具有挑战性的数据集的模拟数据从零度量热仪的ALICE实验在LHC。以及标准基准,如MNIST和CelebA。

英文摘要:

In this work, we propose a novel end-to-end sinkhorn autoencoder with noise generator for efficient data collection simulation. Simulating processes that aim at collecting experimental data is crucial for multiple real-life applications, including nuclear medicine, astronomy and high energy physics. Contemporary methods, such as Monte Carlo algorithms, provide high-fidelity results at a price of high computational cost. Multiple attempts are taken to reduce this burden, e.g. using generative approaches based on Generative Adversarial Networks or Variational Autoencoders. Although such methods are much faster, they are often unstable in training and do not allow sampling from an entire data distribution. To address these shortcomings, we introduce a novel method dubbed end-to-end Sinkhorn Autoencoder, that leverages sinkhorn algorithm to explicitly align distribution of encoded real data examples and generated noise. More precisely, we extend autoencoder architecture by adding a deterministic neural network trained to map noise from a known distribution onto autoencoder latent space representing data distribution. We optimise the entire model jointly. Our method outperforms competing approaches on a challenging dataset of simulation data from Zero Degree Calorimeters of ALICE experiment in LHC. as well as standard benchmarks, such as MNIST and CelebA.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑