- When you did the exercise on cross-validation, what happened to the standard deviation and what does that mean?
The standard deviation stabilizes and reduces on more folds. This means that the performance measurements are more reliable; it is an accurate measure of generalization or overfitting.
- What is the difference between hyperparameters and model parameters?
Model parameters are numerical solutions to a learning algorithm; hyperparameters are what the model needs to know in order to find a solution effectively.
- Is a grid search faster than a randomized search for hyperparameters?
It depends. If the choice of hyperparameters affects the computational complexity of the learning algorithm, then both could behave differently. However, in similar search spaces and in the amortized case, both should finish at about the same time.
- Can I use a regression-based learning algorithm for a classification problem?
Yes, as long as the labels, categories, or groups...
