I don’t remember it very clearly, but I can roughly recall it (0-o)
The question type is still fill in the blanks + judgment + short answer + calculation
Fill in the blanks
I tested the information entropy formula, the five elements of search problems, hingeloss formula, SVM optimization objective function, and the meaning of (X, D, C) in constraint solving problems. There are k balls in an infinite set. What is the distribution of the balls when the entropy is maximum? Time and space complexity of iterative depth-first search. How decision trees prevent overfitting. (Any additions not mentioned are welcome)
True or False
Someone in the course evaluation community provided an outline, which is basically included in the exam outline. Two questions asked whether the Bayesian network is a supervised machine learning method and whether the EM algorithm is an unsupervised machine learning algorithm like.
short answer questions
We consider the similarities and differences between depth-first search and backtracking; the components of the game; the difference between the local algorithm and the global algorithm of the search algorithm, with examples.
Calculation problems
Bayesian network, draw causal diagrams and find simple conditional probability distributions and marginal distributions.
Reinforcement Learning, given a graph, including state, action, transition probability, income, strategy, the first question is to calculate the V value of the state point; the second question is Updated strategy. (You can write it by remembering the update formulas of Q and V)
Neural Network, given a three-layer network diagram, the activation function and loss function of each layer, the first question is to find the weight update formula of back propagation, the second question Substitute the data to calculate the updated weights, and thirdly ask about the impact of the learning rate on training. The network in this question has two nodes in the input layer, three nodes in the hidden layer, and one node in the output layer.
KNN nearest neighbor, give five data points (x1, x2) and their labels, give a data to be classified, first ask where the data to be classified is What categories are classified under different k values? Secondly, ask what the impact of the value of k is.