README.md

Neutrino Particle Direction Predictor

A model based on data from the "IceCube" detector, which observes the cosmos from deep within the South Pole ice. This project is created for the Kaggle Competition IceCube - Neutrinos in Deep Ice

---

Prerequisites

• Dependencies
  • Bash command line
  • Anaconda for python environment management

---

Setup environment

Run the command: source bin/activate.sh at the project root directory.

---

Data

The data source for the project can be found here on the Kaggle page.

It should be placed in the 'data/' directory directly so the file structure looks something like:

• data/
  • test/
  • train/
  • ...

---

Observations of the dataset

• The events have a time element so the observations are dependent on each other.
• Each event has its own set of sub events that are picked up by a sensor
• There is a time series element to this

Tests

Run the following command in the root directory for testing python files

pytest .

Viewing Model stats

This project uses ML flow to monitor model performance.

Use the following command in the root of the folder where the 'mlruns' directory is:

mlflow ui

Trouble shooting

• Error when using Tensorflow when using Nvidia. More Info