Avoiding using incorrect inputs

Hi all,

I am trying to use tensorflow to gather some price data from the web and Im wondering what I should do if one of my requests has a timeout or returns a bogus number.

Just to visualize a simplified system, say I have 5 sources and am polling them repeatedly every 30s and trying to get an average price from them. If one of my sources fails or returns a price 10x higher than it should be I would like to not use this data but also not break my model.

Can I somehow tell my model to not take that input into account? Or what I could do instead?

Thanks for any answers.

Anyone have any tips?