komeisugiura/defn18

Deep Flare Net 2018

Deep Flare Net (DeFN) Astrophysical Journal 2018 version

Komei Sugiura, National Institute of Information and Communications Technology, Japan

0. License

• BSD 3-Clause Clear License

1. Prerequisite

• Ubuntu 16.04 or 14.04
• Python 3.4.3

2. Install

• In the following procedure, ~/work is assumed to be used as a working directory.

$ cd ~/work/
$ git clone https://github.com/komeisugiura/defn18.git
$ cd defn18
$ pip install -r requirements.txt
$ pip install --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.0.1-cp34-cp34m-linux_x86_64.whl

3. Download data

• Visit http://wdc.nict.go.jp/IONO/wdc/solarflare/index.html and download defn_feature_database_v1_pl.zip.

$ cd ~/work/
$ mv defn_feature_database_v1_pl.zip ./
$ unzip defn_feature_database_v1_pl.zip
(password is required)
$ ln -s defn_feature_database_v1/defn_input_database/charval2017X_*.csv.gz ./defn18/data/

4. Run

$ cd ~/work/src
$ ./01RUN_deepflarenet.sh

• The following result will be shown. This means that TSS=0.8024 is obtained by using a pretrained model.

[008000]Acc: Tra=0.8345, Val=0.8584, Tes=0.8584, MaxVal=0.8584(0.8584), TSS=0.8024

5. Training DeFN from scratch

Modify src/deepflarenet.py.

• Uncomment the following line to train the model # net1.train_model(update_interval=100)

• Uncomment the following line to save the trained model. Current model is overwritten. # net1.save_model(myflag.outfile_model)

• Comment the following two lines out, if you don't like to load the model

net1.load_model(myflag.infile_model)
net1.show_training_status(epoch=8000)

A. References

1. N. Nishizuka, K. Sugiura, Y. Kubo, M. Den, and M. Ishii, "Deep Flare Net (DeFN) Model for Solar Flare Prediction", The Astrophysical Journal, Vol. 858, Issue 2, 113 (8pp), 2018. DOI: 10.3847/1538-4357/aab9a7

Local Variables:

coding: utf-8

End: