Error based learning self-assessment questions:
Error based learning self-assessment questions:
- Do I understand the idea of model parameters and a parameterised model?
- Do I understand the idea of an error/cost function?
- Do I understand the idea of an error surface, local and global minima?
- Do I understand the steps taken during the generic learning algorithm?
- Do I understand the ideas related to univariate linear regression?
- Do I understand the ideas related to multivariate linear regression?
- Can I compute the model output, error, sum of error squared, error signal and the new weights given a dataset and a multivariate regression model?
- Do I understand the idea of having to reinitialised weights if a parameterised model gets stuck in a local minima?
- Do I understand the concept of the learning rate?
- Do I understand logistic regression, the expression and how it is applied to create a binary classifier?
- Do I understand the idea of how weights are updated in a similar way to how they were updated for linear regression?
- Do I understand how a logistic regression model is trained and how the weights change over each iteration, resulting in a different model, which could eventually converge?
- Given a logistic regression model and a testing instance, can I determine the prediction?
- Do I understand that categorical features need to be one-hot encoded before applying logistic regression?
- Do I understand why features are normalised before applying logistic regression?
- Do I understand that linear models are not optimal when the data has non-linear relationships?
- Do I understand the idea of transforming a feature via a basis function and how this can result in non-linear models?
- Given a non-linear regression model and a testing instance, can I make a prediction?
- Do I understand the idea that the basis functions can be used in logistic regression thus facilitating non-linear classification models?
- Do I understand how logistic regression can be adapted for multi-class classification?
