Sustainable Development Goals
Abstract/Objectives
This course introduces basic concepts in machine learning and their associated mathematical tools. It is aimed at advanced undergraduates, and assumes no previous knowledge of or machine learning concepts. Knowledge of multivariate calculus, basic linear algebra some familiarity with probabilities would be helpful but not essential. Topics to be covered include probability distributions, supervised learning, Bayesian decision theory, parametric methods, multivariate methods, dimensionality reduction, clustering, nonparametric methods, decision trees, linear discrimination, multilayer perceptron, and deep learning.
Results/Contributions

Students who take this 3-credit course will learn about:

1. Data Analysis – Data preprocessing, visualize data distribution.

2. Build Machine Learning Model – Build different models with the knowledge learned from class, including k-nearest neighbor classifier (KNN), Bayesian classifiers, principal components analysis (PCA), regression model, k-means clustering.

3. Supervised Learning – Understand the differences between supervised learning and unsupervised learning, model selection, and model generalization.

4. Bayesian Decision Theory – By assuming different probability density functions and then implementing Bayesian classifiers through Bayesian theorem.

5. Parametric Methods – Build the parametric model, estimating the model parameters with maximum likelihood estimation. Investigate the number of model parameters influence the performance of classification and regression.

6. Multivariate Methods – Generalize the parametric methods to the multivariate case.

7. Regression – Use gradient descent and least square method to estimate the regression model parameter, and apply this established model to predict interested parameter.

8. Dimensionality Reduction – Why reduce dimensionality? Introduce the concept and implementation of principal components analysis and linear discriminant analysis.

9. Clustering – Introduce k-means clustering, hierarchical clustering, and how to choose the number of clusters.

10. Nonparametric Methods – Introduce the concept of nonparametric methods, and consider it for density estimation, classification, and regression.

11. Linear Discrimination – Introduce the concept of discriminant-based classification, and how to generalize the linear model.

12. Deep Learning – Some basic principle of neural network, and how to use Keras package to implement deep learning.

 

Keywords
ProbabilityClassificationRegression AnalysisSupervised LearningUnsupervised Learning
Contact Information
吳順吉
shunchi.wu@mx.nthu.edu.tw