A tutorial that demonstrates how to predict galaxy M/L ratios using a simple convolutional neural network with the Keras API of TensorFlow 2.0.
This tutorial is written and tested with Python 3.6.
This would be the the preferred option if you already have a local Python environment configured and your machine has a descent GPU.
- Clone this repository to your local environment
- Install and serve a Jupyter notebook host from the cloned directory
- Download the AstroHack datasets to
./data
directory. Unzip any .zip packages there as well. - Run the
astrohack_tf_example.ipynb
notebook and enjoy the fun!
If you are unfamiliar with Python configurations or your local machine does not have a GPU ready for Deep Learning, then you could use a ready-to-go AWS SageMaker notebook instance to get started with.
-
Create an account with AWS if you do not have one yet, and navigate to your AWS console.
-
Select or search for
SageMaker
service in the AWS services section. -
On the navigation panel to the left, find the
Notebook
section and then selectNotebook instances
-
Create a notebook instance
- You can leave most options to their default values
- It is recommend to use
ml.p2.xlarge
orml.p3.2xlarge
instances which allow GPU computation. - In the optional
Git repositories
section, selectclone a public Git repository to this notebook instance only
and enter the URI of this repository (https://github.com/yaodongjia/astrohack-starter-kit-tf.git
). - Confirm and create a notebook instance
- It may take a few minutes before your notebook instance is ready
-
Once the notebook instance is ready, click on
Open Jupyter
and then open theastrohack_tf_example.ipynb
notebook. -
Follow the notebook instructions and have fun!