This is a 8 hours lectures + 8 hours hands-on with tutors.
The Main idea is to give a brief, but useful, introduction to Deep Learning. The particular focus/motivation is Astronomy. If you are not an astronomer do not worry you should be safe. The only thing you might miss is the scientific motivation behind the examples and a deeper understanding on the dataset.
The Main ingredients for running the examples are:
Tensorflow
Keras
anaconda python 3
matplotlib
numpy
Presentations (English only):
Lecture 01
Lecture 02
Lecture 03
Lecture 04
Presentations at XII CBPF School version:
From July 29 to Aug 2 I presented a different version of the minicourse at XII CBPF School. The first talk was given by Dr. Márcio Albuquerque
XII-School-CBPF-01
XII-School-CBPF-02
XII-School-CBPF-03
XII-School-CBPF-04
Video (youtube – portuguese):
Lecture 01
Lecture 02
Lecture 03
Lecture 04
Shareable Examples (Portuguese: Exemplos Compartilháveis)
0. The Simplest example I know
- Strong Lensing Finder (Image Classification)
- Resnet50 Strong Lensing Finder (Image Classification)
- Unsupervised Galaxy Classification from spectra (Auto Encoders and PCA)
- Error with Concrete Dropout
- PSF removal (GAN)
Credit: The examples 1-5 were written by Luciana Olívia Dias and Patrick Schubert under my supervision. There is a similar version of example “0” from adventuresinmachinelearning.com here.
Datasets for the Examples (Portuguese: Dados necessários para os exemplos)
lensdataset for colabs 1 and 2. (Sorry for the zip, wordpress do not allow tar.gz)
spec4000.npz.tar.gz for colab 3
PSF removal GAN for colab 5
For fits image visualization you may use: DS9
To produce beautiful pics from fits you may use: Trilogy
Credit:
The Lens dataset is a preprocessed small portion of the one produced by Bologna Lens Factory Team for the Strong Lensing Challenge.
The galaxy spectra is from sloan, from Astro-ML examples.
The dataset for colab 5 is from arXiv:1702.00403.
