NCTU-ML_2021
This repo is for ML homework in the spring semester studying in NCTU.
Regularized Linear Model Regression (link)
- Apply LSE and Newton method to do regularized linear model regression.
Naive Bayes Classifier and Online Learning (link)
- Apply naive Bayes classifier to classify MNIST handwritten digit.
- Apply online learning to learn the beta distribution of the parameter p of the coin tossing trials
Bayesian Linear Regression (link)
- Generate data from normal distribution
- Sequentially estimating the mean and variance using online learning algorithm.
- Applying Bayesian linear regression to update the posterior and predictive distribution.
Logistic Regression and EM Algorithm (link)
- Apply logistic regression to classify two groups of data.
- Use EM algorithm to classify MNIST handwritten digits.
Gaussian Process and SVM (link)
- Apply Gaussian process to predict the distribution of the data points.
- SVM accuracy:
- linear kernel: 95.08%
- rbf kernel: 98.28%
- linear + rbs kernel: 95.68%
Kernel K-means and Spectral Clustering (link)
- Apply kernal k-means with different intialization method.
- Apply spectral clustering (both ratio cut and normalize cut).
Kernel Eigenface and t-SNE (link)
- face recognition accuracy:
- K-means: 86.67%
- PCA: 90%
- LDA: 76.67%
- t-sne error: 0.8715
- s-sne error: 1.6821