0
$\begingroup$

I'm relatively new to machine learning, and I don't know what error I should use for an RNN.

I want to use a simple Elman RNN to predict the cases of Covid-19 there will be in a hospital for the next 15 days. I modeled this as a regression problem, treating the input like a bunch of dots in a graph to predict the tendency that the data is going to take (only show if there will be more cases or less).
With that bunch of dots I in fact refer to this: enter image description here

Then I would treat this problem as a regression.

I actually don't have anything programmed yet. Firstly I want to write it all on a paper and then get down to work. I am also considering focusing the problem to predict the actual plot of the time-series input, but right now I want to try the regression.

I've come to the conclusion that I can use these four different errors:

  • MSE
  • RMSE
  • Entropy
  • Cross-entropy

What are the different characteristics of these errors? Which to use? Where and when to use them?

$\endgroup$
3
  • 1
    $\begingroup$ Can you clarify this part "treating the input like a bunch of dots in a graph"? Why is the data like that? What is the exact input you're using? $\endgroup$ – nbro Feb 6 at 14:00
  • $\begingroup$ What does the y and x axes on your graph/plot represent? $\endgroup$ – nbro Feb 6 at 18:16
  • $\begingroup$ The x represents the time (in days), and the y is the number of cases in that day. $\endgroup$ – mariogarcia79 Feb 6 at 18:30
1
$\begingroup$

To provide a good answer would fill several pages. To keep it very simple try many different loss functions on your model. Your goal is to have the highest performance based on some desired prediction metric (e.g., RMSE, MAE, MAPE, etc.). You almost always have plenty of time to try many loss functions so you don't need to have a full understanding, and few people do, to start your project.

I recommend you read the following to learn more:

$\endgroup$
3
  • $\begingroup$ Thank you, the second page was really useful and clear. Now I understand better this topic $\endgroup$ – mariogarcia79 Feb 6 at 17:40
  • 1
    $\begingroup$ @mariogarcia79 You should still clarify what I asked under your post, if you want to get a more specific answer. At this point, your question has nothing to do with RNNs, but you're just asking when to use MSE as opposed to CE. $\endgroup$ – nbro Feb 6 at 18:01
  • $\begingroup$ @nbro This is what I was refering to. $\endgroup$ – mariogarcia79 Feb 6 at 18:10

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.