0
$\begingroup$

I have a baseline model which has certain hyperparameters to tune (it's actually a neural network, but I don't know if it's important in this context). I want to compare it with my own extension of the baseline model, which is built on top of it and has some additional hyperparameters to tune.

Should I tune baseline parameters when evaluating the baseline and tune both types of parameters (baseline + specific for the extension) when evaluating my new model? Or should I "freeze" the best set of baseline parameters taken from the first evaluation when evaluating the new model and tune only the specific ones?

$\endgroup$
0
$\begingroup$

The concept of a baseline model is that it is the simplest way to solve a problem. In some cases we may use 0 rules(0R, Random Guessing) model as a base line.

So we can judge and evaluate the next trials that uses more complex models(Like neural networks) as an improvement in accuracy, precision,..etc than the base line model.

from this concept you should start building a base line model with the minimal fine-tuning that will help you evaluate you trials with respect to

  1. Interpretation Complexity.
  2. Predictive power.(you can map it to accuracy in the most cases)
  3. Time to build.
  4. Memory consumption.
$\endgroup$

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.