1. (20 points) Answer whether the following statements are True or False. (a) (2
ID: 3870320 • Letter: 1
Question
1. (20 points) Answer whether the following statements are True or False.
(a) (2 point) Parameters of a multivariate linear regression problem can be expressed as a closed form equation.
(b) (2 point) Parameter estimation using MLE and MAP techniques always produce identical result.
(c) (2 point) In a univariate regression problem, choice of polynomial regression with (high degree polynomial) is always preferred over linear regression.
(d) (2 point) An overfitted model ensures good generalization ability.
(e) (2 point) Either in classification or in regression problem, reducing bias always results in reduced variance.
(f) (2 point) In case of a binary classification problem where each class is modeled as a multivariate Gaussian distribution, the decision boundary that separates the two classes is always linear.
(g) (2 point) Number of parameters that need to be estimated for a linear regression in univariate case is 2.
(h) (2 point) In parameter estimation using Maximum Likelihood Estimate technique, any prior information/knowledge of the parameter is used.
(i) (2 point) In the univariate case, when we use parametric classification and where class conditional probabilities are modeled as Gaussian distribution, the decision region of any class is always contiguous and can never be disjoint.
(j) (2 point) Inverting a diagonal covariance matrix takes less time as compared to a full covariance matrix
Explanation / Answer
1) Parameters of a multivariate linear regression problem can be expressed as a closed form equation.
Ans: True
2)Parameter estimation using MLE and MAP techniques always produce identical result.
Ans: False. They procuce approximates.
3) In a univariate regression problem, choice of polynomial regression with (high degree polynomial) is always preferred over linear regression.
Ans: True.
4)An overfitted model ensures good generalization ability.
Ans: True.
5)Either in classification or in regression problem, reducing bias always results in reduced variance.
Ans: False. It's not mandatoey
6)n case of a binary classification problem where each class is modeled as a multivariate Gaussian distribution, the decision boundary that separates the two classes is always linear.
Ans: True.
7) Number of parameters that need to be estimated for a linear regression in univariate case is 2.
Ans: True.
Related Questions
Navigate
Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.