/ohw18_omlet

Oceanhackweek 2018 Ocean Machine LEarning Toolkit

Primary LanguageJupyter Notebook

ohw18_omlet

Oceanhackweek 2018 Ocean Machine LEarning Toolkit

Ocean Machine LEarning Toolkit "OMLET"

Participants

  • Drew Snauffer
  • Manuel Valera
  • Ben Larson
  • Candice Hall
  • Kenneth Jackson
  • Spencer Sherk
  • Xiangming Zeng
  • Joseph Gum

The problem

To make an OMLET you need to crack some data. We propose to make machine learning models that have been trained on shipboard data to explore the data around OOI installations Station Papa, Coastal Pioneer, and Coastal Endurance. We will be start with the following parameters:

  • Temperature
  • Salinity
  • Dissolved Oxygen
  • Chlorophyll

In order to eggsecute the hack we will use different toolkits, such as:

  • Keras
  • SciKit-Learn

Application Example

We want to take shipboard data, such as a (sub)set of Line P data for Station Papa, and train the network to recognize good data vs anomalous data. Anomalous data can represent a problem with the sensor to an interesting phenomena different from the norm.

Specific Tasks

  1. Create a simple format for data to be fed into the model
  2. Build ML models
  3. Testing to see models work
  4. Cross-validation

Existing methods

Linear regression(s)

Proposed methods/tools

Machine learning

Background reading

Optional: links to manuscripts or technical documents for more in-depth analysis.