0
$\begingroup$

I'm trying to solve a problem where I need to train one model with N dimensions and again train on top of that model with M dimensions. How can I achieve it?

To give you guys some context, I have 1 bluetooth beacon talking to 5 bluetooth readers that give out some value. Based on the value given out by the reader, I'm predicting the location of the beacon. My training data has 5 dimensions because there are values from 5 readers.

Now I have another scenario where I have 10 readers instead of 5. How do I train on top of my first model with 10 readers this time? Is it even possible?

The number of readers varies depending on the size of the room. So there's no way I can tell that there will always be a fixed number of readers (that is, fixed number of dimensions).

How do I go about solving this problem? Any help is appreciated!

$\endgroup$
0
$\begingroup$

You could try to summarise the data before handling it to the model:

If you have 5 readers, you could gather the information of "mean", "max", "min", "standard deviation" of them, this will work on 10, 15, any N of readers. Unless there is something which stops you from calculating mean, max, etc.

| improve this answer | |
$\endgroup$
  • $\begingroup$ I need to give you more context to the problem to explain why I can't take the mean, max, min etc. The readers give out a value called RSSI value. The closer the beacon is to the reader, the higher the RSSI value is. So based on these values, my model is learning to predict the location of the beacon. The model will lose a lot of information to learn from if we take the mean, min, max or standard deviation. The model needs to know how strong or weak the RSSI value from each reader is in order to predict the location accurately. $\endgroup$ – Shreyas S Apr 25 '19 at 4:53
  • $\begingroup$ You're right!! I had to redefine my problem statement though. I'm taking the mean value right now and running a few experiments to see how it turns out. Thank you so much! :) $\endgroup$ – Shreyas S Apr 29 '19 at 9:37

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.