search

nipunbatra / transferable-energy-breakdown-old

1 stars 0 forks source link

Implementing parameter transformation across regions #11

Open nipunbatra opened 7 years ago

nipunbatra commented 7 years ago

Yiling, PS my mail titled: Learning with priors (params from other regions)

This code should help you get started. With similar concepts, you would need to implement the basic transformation strategies for appliance latent factors.

We can have various such strategies. For starting, let us use:

  1. A_sd = alpha \times A_au + beta
  2. A_sd = [[alpha 1, alpha 2, ...], [alpha4,....]] \times element wise A_au + [[beta 1, ...]]

Again, we don't want to analyse the results expect for sanity checking. Once I have done the data cleaning, we can then use this implemented code.

yilingjia commented 7 years ago

Hi Nipun,

I tried to implement the transfer learning in "base-transfer-Yiling.ipynb". This is the first version and I will check with professor Wang tomorrow to see if that is correct.

As I use different learning rates and iteration number, the rmse seems get smaller than just use A_au directly except hvac.

I will check this tomorrow.

Thanks!