行星探索数据

行星亮度数据集，人工智能机器学习项目，循环神经网络解决，CNN+RNN，为地球人寻找一个居住星球一直是引人注目的话题,以往从数以万计的太阳系外行星检测过境行星只能依靠人的肉眼去辨识,极为低效.现在通过机器学习和人工智能方法能够帮助我们高效地进行行星探索任务,Kyle A. Pearson等人在最近的工作中,发表了成功应用一维卷积神经网络模型(1D-CNN) 在模拟数据上识别凌星.本文从影响模型性能的数据集和1D-CNN 的卷积核配置这两方面对此工作做进一步探讨.结果显示,与原来精度88.45%相比: 通过优化卷积核的配置以及分别用光变曲线参数离散,光变曲线参数连续的数据集作为1D-CNN模型的训练集和测试集,1D-CNN模型识别凌星的精度能够提升2%—3%.

Planetary Brightness Dataset is an artificial intelligence and machine learning project utilizing CNN+RNN (Convolutional Neural Network + Recurrent Neural Network). The search for habitable planets for humanity has long been a prominent research topic. Previously, detecting transiting exoplanets from tens of thousands of candidates relied exclusively on manual visual identification, which was highly inefficient. Currently, machine learning and artificial intelligence approaches can significantly enhance the efficiency of planetary exploration missions. In a recent work, Kyle A. Pearson et al. successfully applied a one-dimensional convolutional neural network (1D-CNN) model to identify transiting exoplanets using simulated data. This paper further investigates this prior work from two aspects: the dataset factors impacting model performance and the convolution kernel configuration of the 1D-CNN. The results show that, compared to the original accuracy of 88.45%, the identification accuracy of the 1D-CNN model for transiting exoplanets can be improved by 2% to 3% by optimizing the convolution kernel configuration and using datasets with discrete light curve parameters and continuous light curve parameters as the training and test sets of the 1D-CNN model respectively.

搜集汇总

背景与挑战

背景概述

该数据集包含行星亮度数据，用于人工智能机器学习项目，旨在通过一维卷积神经网络（1D-CNN）优化来识别凌星（过境行星）。基于Kyle A. Pearson等人的研究，通过调整卷积核配置并使用光变曲线参数离散与连续的数据集进行训练和测试，模型识别精度从88.45%提升了2%—3%。