AstroHack - Tensorflow + Keras Example

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.

How to Use

Local Environment

This would be the the preferred option if you already have a local Python environment configured and your machine has a descent GPU.

1. Clone this repository to your local environment
2. Install and serve a Jupyter notebook host from the cloned directory
3. Download the AstroHack datasets to ./data directory. Unzip any .zip packages there as well.
4. Run the astrohack_tf_example.ipynb notebook and enjoy the fun!

Using a AWS SageMaker Notebook Server

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.

1. Create an account with AWS if you do not have one yet, and navigate to your AWS console.

2. Select or search for SageMaker service in the AWS services section.

3. On the navigation panel to the left, find the Notebook section and then select Notebook instances

4. Create a notebook instance

   • You can leave most options to their default values
   • It is recommend to use ml.p2.xlarge or ml.p3.2xlarge instances which allow GPU computation.
   • In the optional Git repositories section, select clone 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

5. Once the notebook instance is ready, click on Open Jupyter and then open the astrohack_tf_example.ipynb notebook.

6. Follow the notebook instructions and have fun!