Machine Learning Engineer Nanodegree
Capstone Proposal
Ricardo Castellanos
January 8, 2021
The included Jupyter Notebook was run under an Amazon Sagemaker instance. I set up an Amazon Elastic Compute Cloud (EC2) instance on AWS to run Jupyter Notebook on GPU. I have increased service limits for p2.xlarge instance type (by default it’s 0).
This instance included CUDA support. If you are not using GPUs you will need to replace the CuDNNLSTM calls with basic LSTM calls.
The code runs with kernel conda_tensorflow_p27, and requires the following libraries, in no particular order
tensorflow
keras
numpy
scipy
scikit-learn
matplotlib
Data files required to run the code are the exoTest and exoTrain labelled time-series datasets from https://www.kaggle.com/keplersmachines/kepler-labelled-time-series-data.
Or download the compressed csv from the repo exoTest.csv.zip and exoTrain.csv.zip