regn:使用神经网络对全球降水进行稳健估计

  • i1_842581
    了解作者
  • 23.9MB
    文件大小
  • zip
    文件格式
  • 0
    收藏次数
  • VIP专享
    资源类型
  • 0
    下载次数
  • 2022-05-09 11:37
    上传日期
使用神经网络(REGN)可靠地估算全球降水 首字母缩略词REGN代表使用神经网络对全球降水进行稳健估计。 同时,瑞格([rɛŋn])是瑞典语中的雨。 REGN项目的目的是开发基于神经网络的GPROF算法实现。 该存储库用于收集该项目的所有代码和结果。 AGU介绍 REGN项目的中间结果已在2020年AGU的演讲中介绍 H206-07-使用神经网络从GPM被动微波观测中获取贝叶斯降水 作为会议H206的一部分-天基降水观测和估算:科学和应用的创新I。 演示文稿的幻灯片可以在找到。 运行代码 重现显示的结果所需的代码由两部分组成: regn Python程序包,该程序包实现基于QRNN的GPROF检索 和文件夹中包含的,其中包含执行数值分析的Python代码。 Python依赖项 注意:在安装任何这些依赖项之前,最好使用Python venv或conda创建一个新环境。 我们的工作基于
regn-master.zip
内容介绍
# Robust Estimation of Global precipitation using Neural networks (REGN) The acronym REGN stands for *Robust Estimation of Global precipitation using Neural networks*. At the same time, *regn* ([rɛŋn]) is the swedish word for rain. The aim of the REGN project is to develop a neural-network based implementation of the GPROF algorithm. This repository is used to collect all of the code and and results from this project. ## AGU Presentation Intermediate results from the REGN project have been presented at AGU 2020 in the presentation > H206-07 - Using Neural Networks for Bayesian Precipitation Retrievals from GPM Passive Microwave Observations as part of the session *H206 - Space-Based Precipitation Observations and Estimation: Innovations for Science and Applications I*. Slides from the presentation can be found [here](https://raw.githubusercontent.com/SEE-MOF/regn/master/presentation/agu_presentation.pdf). ## Running the code The code required to reproduce the presented results consists of two parts: - The ``regn`` Python package, which implements the QRNN-based GPROF retrieval - The Jupyter notebooks contained in the [notebooks/gmi](notebooks/gmi) and [notebook/mhs](notebooks/mhs) folders, which contain the Python code which performs the numerical analyses. ### Python dependencies > **Note**: Before installing any of these dependencies it is probably a good > idea to create a new environment using Python venv or conda. Our work builds on and requires a range publicly available packages, which are collected in the ``requirements.txt``. After cloning this repository, you can install these packages using: ```` $ python3 -m p install -r requirements.txt ```` ### Installing the ``regn`` package To run any of the notebooks, the ``regn`` package must be in your ``PYTHONPATH``. The easiest way to achieve this is probably to just install the package using ``pip``: ```` $ python3 -m p install -e . ```` ## The QRNN implementation We have recently migrated our implementation of QRNNs from the [typhon](https://github.com/atmtools/typhon/) package to a new, separate package called [quantnn](https://github.com/simonpf/quantnn). This is still relatively new and lacks extensive documentation but is what has been used within this study. ## References For background information on quantile regression neural networks (QRNNs), refer to the following article: - Pfreundschuh, S., Eriksson, P., Duncan, D., Rydberg, B., Håkansson, N., and Thoss, A.: A neural network approach to estimating a posteriori distributions of Bayesian retrieval problems, Atmos. Meas. Tech., 11, 4627–4643, [https://doi.org/10.5194/amt-11-4627-2018](https://doi.org/10.5194/amt-11-4627-2018), 2018. For more information on the current GPROF algorithm, please refer to the following publications: - Kummerow, C. D., Randel, D. L., Kulie, M., Wang, N. Y., Ferraro, R., Joseph Munchak, S., & Petkovic, V. (2015). The evolution of the Goddard profiling algorithm to a fully parametric scheme. Journal of Atmospheric and Oceanic Technology, 32(12), 2265-2280. - [The GPROF version 5 ATBD](http://rain.atmos.colostate.edu/ATBD/ATBD_GPM_June1_2017.pdf)
评论
    相关推荐
    • RNN.zip
      基于梯度下降的RNN,训练过程用到了BPTT算法,通俗易懂
    • 捷联惯导积分算法_SAVAGE
      捷联惯导经典算法姿态更新以及速度和位置更新 包括英文原文及其纠错
    • template-scala-rnn:RNN算法实现
      该模板提供情感分析算法 。 安装。 遵循。 安装后启动所有 PredictionIO 供应商并检查 pio 状态: pio-start-all pio status 将此模板复制到您的本地目录: pio template get ts335793/template-scala-spark-...
    • lip-movement-net:使用基于唇动的RNN检测器进行说话人检测
      RNN的检测算法如下: 从输入视频文件或网络摄像头中接受25帧的序列(即25 fps视频的1秒钟的帧值)。 序列长度是可配置的。 对于此序列中的每个视频帧: 将其转换为灰色框。 这样可以加快接下来的几个步骤(减少...
    • Algorithm-RNN-SM.zip
      Algorithm-RNN-SM.zip,RNN-SM:使用递归神经网络T-IFS对VoIP流的快速隐写分析,算法是为计算机程序高效、彻底地完成任务而创建的一组详细的准则。
    • vec2.js:操纵2D向量
      另外,由于Infinity和NaN非常阴险,因此该库将在检测到这两个库后立即抛出该异常,因此您可以采取措施来修复数据/算法。 支持的运营 变更([fn]) 添加一个观察者fn ,只要此向量改变,该观察者就会被调用。 在...
    • CMOS校验和算法
      CMOS校验和算法~~~!~~!~~!~!`1
    • rnn.rar
      逆向最近邻查询算法在实际应用中有很多领域
    • rnn-by-numpy:使用numpy实现rnn和语言模型
      本仓库主要用numpy从头开始构建rnn结构,包括前向传播算法 反向传播算法 学习率 随机梯度下降; 实例:给定一个x来预测y,虽然这个没有实际意义,所以在这里主要目的是为了阐释rnn算法 x: SENTENCE_START what are ...
    • RNN算法打包matlab
      matlab中rnn的用法