Open JackKelly opened 3 years ago
optimistically speaking, approach 1 "should" be handled by dropout on the input layer. As long as the missing values are set to the same as the training-time dropout uses (presumably zeroes). But that's pure optimism
hehe, I like optimism (and that's a much simpler approach to implement!) I'll try that first. Thanks for the comment! I hope all's well :)
UPDATE: I've started a separate issue to remind me to try dropping out PV metadata: #48
Two approaches:
1) One network which gets PV metadata when it's available. When it's not available, somehow mask those inputs. Set to -1? Or have a separate 'mask' input?
2) Two networks: One which predicts distribution of PV yield, without knowing any PV metadata. A second network which takes that PV distribution, and refines it when metadata is available