If you have a logistic regression model and the size of the model is too big. What would you do to decrease the size of the model?
Please note that you are free to retrain the model from scratch.
[This is an interview question. Please feel free to post your answer below.]
Dimensionality reduction any method we use like backward elimination or we can check and find correlated features we can apply PCA ?
The PCA model will require decent space. So, with PCA, the size of the model will actually increase.
Could you please elaborate a bit more on what do you mean by the size of the model?
The model is a bunch numbers. Right?
In Machine learning, we try to figure out “Function”.
Input -> Function -> Output
If the function is a decision tree, the model is basically the tree of decision boundaries.
And in case of the function being regression, the model is basically list of theta.
t0 + t1f1 + t2f2… => result
f1, f2, f3 are the input features.
t0, t1, t2 are the numbers that constitutes the model. We need to find out the values of t0, t1, t2 as part of training.
I hope that makes it clear.