论文标题
使用神经普通微分方程的星系形态分类
Galaxy Morphology Classification using Neural Ordinary Differential Equations
论文作者
论文摘要
为了获得星系形态分类的目的，我们介绍了称为神经常见微分方程（节点）的残留网络（RESNET）的连续深度版本。我们从Galaxy Zoo 2数据集中对星系图像进行分类，该数据集由五个不同的类别组成，并根据图像类别的不同，获得了91-95 \％之间的精度。我们使用不同的数值技术训练节点，例如伴随和自适应检查点（ACA），并将它们与重新连接进行比较。尽管Resnet具有某些缺点，例如耗时的体系结构选择（例如，层数）以及培训所需的大数据集的要求，但节点可以克服这些限制。通过我们的结果，我们表明节点的准确性与RESNET相当，并且与Resnet相比，所使用的参数数量约为三分之一，从而导致记忆足迹较小，这将使下一代调查受益。
We introduce a continuous depth version of the Residual Network (ResNet) called Neural ordinary differential equations (NODE) for the purpose of galaxy morphology classification. We carry out a classification of galaxy images from the Galaxy Zoo 2 dataset, consisting of five distinct classes, and obtained an accuracy between 91-95\%, depending on the image class. We train NODE with different numerical techniques such as adjoint and Adaptive Checkpoint Adjoint (ACA) and compare them against ResNet. While ResNet has certain drawbacks, such as time consuming architecture selection (e.g. the number of layers) and the requirement of a large dataset needed for training, NODE can overcome these limitations. Through our results, we show that that the accuracy of NODE is comparable to ResNet, and the number of parameters used is about one-third as compared to ResNet, thus leading to a smaller memory footprint, which would benefit next generation surveys.
