From what I've seen, neural networks take a set of atomic inputs.

I want an input to be a variable array, i.e. a group of people (with unique IDs).

If I didn't care about their ID, I could simply feed a count of people as an input - but I do.

My thought are to have some form of forward-feeding, and iterate through each person, using their ID as an input to a node representing "insert entered room"

Does this make sense?

The same ID would be used in different nodes at different points. I.e to a "user left room" or "user sat down"

Also, how would one model two users interacting, atomically for inputs? I'm guess use 2 inputs representing "interacter A" and "interacter B"

  • $\begingroup$ You can use maximum size array and fill empty cells with zeros. Alternatively you can use RNN and augment input with permutations (if items are unordered) $\endgroup$ May 12 '19 at 6:07
  • $\begingroup$ It sounds to me that you're using the wrong tool. Your id numbers are effectively labels, not numerical values. There is presumably no relation between people with similar id values, but to the NN this would be the case. $\endgroup$ May 12 '19 at 9:00
  • $\begingroup$ @OliverMason I see what you mean. How would you model this scenario? $\endgroup$
    – Tobi
    May 12 '19 at 16:56
  • $\begingroup$ Depends what you want to do with it! $\endgroup$ May 12 '19 at 22:04
  • $\begingroup$ @OliverMason perform actions based on actions other users have taken. I need the NN to keep track of users $\endgroup$
    – Tobi
    May 12 '19 at 22:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.