1. What are the reasons of model overfitting with respect to a decision tree classifier? If you suspect that your decisi
Posted: Fri May 20, 2022 2:55 pm
1. What are the reasons of model overfitting with respect
to a decision tree classifier? If you suspect that your decision
tree model is overfitted, how would you fix the model?
2. Why do you need to use an appropriate combination of
activation and loss function for Deep Neural Network models? Give
example of an appropriate pair of activation and loss function.
3. Give an example of an ensemble method that manipulates
the training set and perform iterative modeling by adaptively
changing the distribution of training examples selected for the
base detectors to learn the model. Explain how the detector you
mentioned above works.
4. Suppose a bank wants to develop a classifier that guards
against fraudulent credit card transactions. Illustrate how you can
induce a quality classifier based on a large set of nonfraudulent
transactions and a very small set of fraudulent transactions.
Giving an explanation of just one approach will suffice.
5.
For each of the Boolean expressions given below, state whether
the problem is linearly separable or not. Giving yes.no answer for
each expression will suffice.
(a) A AND B
(b) A OR B
(c) NOT A AND B
Hint: You can draw the truth table and a two-dimensional plot
for each of the Boolean expressions to figure out the answers. The
XOR function plot might help from problem 21 (short
answer).
to a decision tree classifier? If you suspect that your decision
tree model is overfitted, how would you fix the model?
2. Why do you need to use an appropriate combination of
activation and loss function for Deep Neural Network models? Give
example of an appropriate pair of activation and loss function.
3. Give an example of an ensemble method that manipulates
the training set and perform iterative modeling by adaptively
changing the distribution of training examples selected for the
base detectors to learn the model. Explain how the detector you
mentioned above works.
4. Suppose a bank wants to develop a classifier that guards
against fraudulent credit card transactions. Illustrate how you can
induce a quality classifier based on a large set of nonfraudulent
transactions and a very small set of fraudulent transactions.
Giving an explanation of just one approach will suffice.
5.
For each of the Boolean expressions given below, state whether
the problem is linearly separable or not. Giving yes.no answer for
each expression will suffice.
(a) A AND B
(b) A OR B
(c) NOT A AND B
Hint: You can draw the truth table and a two-dimensional plot
for each of the Boolean expressions to figure out the answers. The
XOR function plot might help from problem 21 (short
answer).