2024 Fall Statistical Models Final Exam Recap
"TLDR: This article reviews the content of the final exam for the 2024 Fall Statistical Models course, which includes two parts: short-answer questions and calculation problems. The short-answer questions cover a comparison between C4.5 and ID3, an introduction to the EM algorithm, the implementation steps of a term-document matrix, common NLP tasks and feature applications, the concept and methods of clustering, the role and differences of dataset splitting, as well as probability distribution functions. The calculation section requires students to master knowledge points such as deriving the likelihood function for a normal distribution, applying the K-means algorithm, solving template problems for the Naive Bayes classifier, and using n-gram models to compute sentence probabilities."
htk is a nice person; the questions were quite simple, and they even couldn't bear to make everyone do two rounds of k-means.
Short Answer Questions
- Improvements of C4.5 compared to ID3
- Briefly introduce the EM algorithm
- Introduce the term-document matrix and the steps to construct this matrix
- Introduce common NLP tasks and feature applications
- What is clustering? Give 2 common clustering methods
- The roles and differences of training set, validation set, and test set
- Write the probability density function of the Poisson distribution and the meaning of its parameter
- Write the probability density function of the normal distribution
Calculation Questions
- Write the likelihood function of the normal distribution and use maximum likelihood estimation to derive
- Given 6 data points, use K-means to compute one round and find the centroids of the two clusters
- A template problem on Naive Bayes
- Use n-gram to calculate the probability of a sentence (n = 2)
- Given two sentences, use the bag-of-words model to produce word vectors