DS5220: Supervised Machine Learning
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
DS5220: Supervised Machine Learning
Final review
Final structure
Final will be cumulative but mostly focus on the second half of the semester.
35 multiple choices questions
2 calculation questions
3 coding questions, need to show the code
Concepts review
KNN
The algorithm of KNN
The advantages and disadvantages of KNN
distance metrics
The effect of K in KNN
SVM
Definition of hyperplane (including the formula)
Definition of margin and support vectors
The function of soft margin and slack variable
The function of kernel
When to use each type of kernels?
What is kernel trick?
Maximal margin classifier → Support vector classifier → Support vector machine
Comparing SVM with other classifiers
SVM with more than two classes
Difference between SVR and SVC (ϵ-tube)
Difference between SVR and linear regression
Decision tree
The structure of the decision tree
Greedy algorithm
Recursive binary splitting
The effect of tree size
Cost complexity pruning
Definition of Gini index and entropy and when to use
Comparison between SVM and decision tree
Advantages and disadvantages of decision tree
Ensemble learning
Advantages and disadvantages of ensemble
How to achieve diversity in ensemble learning
Comparison between bagging and boosting (Training, Diversity, Focus)
Bagging
Definition for bagging
Example: Random Forest
Bootstrap sampling
Out-of-bag error
Randomness in Random Forest
How to decide the result in Random Forest
When to use Random forest
What will happen if the number of trees is too large/too small?
What does variable importance plot do in the Random Forest?
Boosting
Definition for boosting
Example: Adaboost
When boosting will fail?
Other type of boosting
Neural network
Definition of nodes and links
When we want to use the hidden layer
Difference between activation level and activation function
Difference and connection between the Perceptron and SVM
Feed-forward neural network
The hyper parameters in FNN
Choice of activation function for various type of questions
The process of backpropagation
Batch normalization, Regularization and Dropout
Convolution neural network
Convolution operation
Padding
Calculate the size of outputs and number of parameters
Recurrent neural network
When to use RNN
When RNN may fail
2023-12-13