DS807: Applied machine learning

The students are expected to be familiar with the following topics:

• Regression analysis.
• Basic unsupervised methods, including principal component analysis and clustering.
• Sampling techniques such as the Bootstrap, cross-validation, and data split (train and test), and how (and when) they should be applied.
• Basic understanding of “modern” approaches to statistical learning, including decision trees.
• Application of the above techniques in a programming language.

This knowledge can be obtained through the courses DS804: Data mining and machine learning; DS805: Multivariate Statistical Analysis and, if possible, DS810: Data Driven Decision Making.

The aim of the course is to enable the student to apply the most commonly used methods in machine learning.

The focus of the course to:

• Give the competence to setup a complete applied machine learning analysis from beginning to end.
• Give skills to perform classification and predictions using statistical and deep learning methods and to make critical assessments of the results.
• Give knowledge and understanding of a broad range of machine leaning techniques and to access their strengths and weaknesses when applied to data of varying types.

The learning objective of the course is that the student demonstrates the ability to:

• Develop an understanding of the fundamental concepts of machine learning, including algorithms, models and practices.
• Implement good methods and practices for effective deployment of machine leaning systems
• Acquire practical competencies and hands-on experience in applying machine learning methods in quantitative and qualitative workflows using a variety of data types.
  

The following main topics are included in the course. Under each topic presented the main focus will be on the applications of the models and methods:

•    Classical statistical learning tools and their applications (shallow learners):

1. Support vector machines
2. Decision trees, boosting, random forests, and gradient boosting
3. Ensembling

•    Deep Learning Methods and their applications:

1. Fully connected neural networks 
   
2. Convolutional neural networks
3. Recurrent neural networks
4. Training, regularization and optimization
5. Autoencoders and variational autoencoders
6. GANs
7. Transformers and Generative AI

Models, applications and data will be inspired by cases posted on Kaggle and Hugging face.

• Deep Learning with Python by Francois Collet (ISBN10: 9781617294433)

See itslearning for syllabus lists and additional literature references.

Indicative number of teaching hours covers more than 45 hours per semester, divided into llectures and exercises. 

Activities during the study phase:

• Attending lectures, including computer lab session
• Solving ad hoc assigments
• Self study of various parts of the course material.