Revision as of 00:31, 9 July 2016 by ARuiz (Talk | contribs)

Jump to: navigation, search

Week 1:

  • Read Superforecasting: The art and science of prediction. Random House, 2015 by Tetlock, Philip, and Dan Gardner.
  • Met with mentor for background on the GasDay Labs mission and our research topic.
  • Met with mentor for background on the specifics in GasDay Labs models.
  • Started researching the scientific literature for examples of uses of probabilistic forecasting.
  • Worked with the Brier score metric in order to identify potential flaws.

Week 2:

  • Further reading into applications for probabilistic forecasting.
  • Implemented basic probabilistic forecasting scoring metrics (pinball, Winkler, Brier).
  • Read papers on probabilistic load forecasting.
  • Familiarized myself with GasDay Labs' database.

Week 3:

  • Attended ethics training
  • Started learning about machine learning by implementing simple multivariate linear regression algorithms.

Week 4:

  • Learned about categorical classification machine learning algorithms.
  • Read papers about about different strengths and weaknesses in scoring metrics.
  • Prepared presentation for GasDay talk.

Week 5:

  • learned about regularization for some machine learning models.
  • Read uncertainty quantification papers.
  • Read papers on probabilistic forecasting.
  • Prepared and practiced presentation for REU talk.

Week 6:

  • Started learning about the implementation of neural networks.
  • Read uncertainty quantification papers
  • Finally finished reading Gneiting et al. (2014) paper on probabilistic forecasting.
  • Interviewed engineers for insight on uncertainty quantification.