search

tholoien / empiriciSN

Generate realistic parameters for a SN given host galaxy observations based on empirical correlations from SN datasets
MIT License
2 stars 4 forks source link

Train Model #12

Closed tholoien closed 8 years ago

tholoien commented 8 years ago

Split the SN sample into training and validation sets and use this to train the model. Use cross-validation to determine the right number of parameters to incorporate and measure the accuracy of the model.

drphilmarshall commented 8 years ago

When we met with Risa, we discussed the following splits:

  1. No split: all SNe analysed as a single population
  2. Exp hosts vs DeV hosts
  3. High z vs Low z
  4. SDSS vs SNLS

To assess significant differences between the resulting models, we'll need a way of quantifying uncertainty in the GMM parameters: we decided we'd need jackknife or bootstrap, if we go with the astroML XDGMM...

On Fri, Jul 8, 2016 at 3:38 PM, tholoien notifications@github.com wrote:

Split the SN sample into training and validation sets and use this to train the model. Use cross-validation to determine the right number of parameters to incorporate and measure the accuracy of the model.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/tholoien/empiriciSN/issues/12, or mute the thread https://github.com/notifications/unsubscribe/AArY95-zf-lVJl1v-gjO2iF2Dg1u0JNVks5qTtFhgaJpZM4JIc76 .

drphilmarshall commented 8 years ago

Looks like this issue can be closed, @tholoien ! You can re-issue the CV optimization separately, but for now you have a trained model that can be used to explore the conditional PDFs to allow SNe to be predicted in new (test set) host galaxies.