Frosi Exam: Chapter 2

Was auf diesem Rad steht (39 Optionen)

• What is the idea behind the Linear Regression model? Given the set of variables, how can you solve the linear regression problem?
• What is the purpose of regression? How would you check if the regression line is the best? How do you evaluate it? How would you predict the value? What is the equation?
• What equation would you use to check the goodness of the models?
• What if we had 1000 training data and 100 models, would you use them all?
• Lasso regression. What is the form? What does lasso try to minimize? How so?
• How can you solve non-linear regression? Write the model.
• How would you fix multicollinearity?
• How do you expect the residual plot to be? What are the main features of residual plot? How are they distributed? What does the mean value of the residues equal to?
• Write the residue sample square formula? Suppose instead you have 2 features? How would you change the equation?
• In regularization, what is lambda?
• Write the residual sum of squares error formula. Standard form (linear model). Write the same formula but with 2 features and change the residual sum of square errors.
• What is the name of the phenomenal that happens when you have the noise of a linear or non-linear model which is biased to the data itself.
• What is polynomial regression? Give an example with 2 features (write it out).
• What is the equation for multivariate linear regression? Write the RSS too.
• What is the Moore-Penrose pseudo-inverse? Why is it needed?
• What happens when our data shows non-linear behavior?
• What does multivariate linear regression assume?
• What is the non-linear linear regression model? Write the equation?
• What is the Residual Standard Error? Write the equation.
• What is the total sum of squares? What does it measure?
• What is the coefficient of determination? What does it measure?
• Models should evaluate data that has not been used to build the model. What are the types of data?
• What is Hold-Out Evaluation?
• What is the problem with too small training set?
• What is the problem with too small test set?
• What is cross-validation?
• What is synergy? How can we translate the non-linearities to the usual linear model?
• What is the phenomenal that occurs when we have interactions between variables?
• What would a “good” residual plot look like?
• How would we transform non-linearities?
• How can we identify outliers in a residual plot?
• What is the effect heteroscedasticity?
• What are high leverage points? Why are do they pose a problem?
• Points with studentized residuals greater than what number in absolute value may be considered an outlier?
• What is collinearity? How do we solve this?
• How do we improve linear regression?
• What are some alternatives to remove unnecessary features?
• What are the 2 shrinkage methods? When would we use them?
• What needs to be done before learning a ridge regression model?