View All Pages

Examination - Learning and Knowledge

The examination for this module is divided into three parts. The first two are hands-on assignments where you get to work with machine learning tools and the third one is aimed at exploring the area.

Given that machine learning is such an important tool in the toolbox of an engineer and researcher of today, you will likely enter this course with very different background knowledge. To cater for this we will allow you suggest variations of the first two assignments as long as you can show that they fulfil the same objectives as the original ones. We hope that some of you will take the opportunity to explore the tools with your own research domain or maybe take on a Kaggle challenge instead. We want a proposal latest Sept 24 that describes what you will do and which of the assignments 1 and/or 2 it replaces and how. We will judge your proposal based on

    • You have the background to do it
    • A realistic project (remember you have 16h or 32h depending on if you want to replace one or both of the assignments).
    • Clear goals and ways to tell if you meet these goals
    • Cooperation between several people

 

Assignment 1: Hands-on supervised learning

Objectives

  • Hands-on experience with an ML tool such as scikit-learn
  • Applied non-deep supervised learning methods such as 
    • logistic regression
    • discriminant analysis: LDA, QDA
    • k-nearest neighbour
    • tree-based methods: classification trees, random forests, bagging
    • boosting
    • SVM
  • Understanding of how raw data can be turned into a form suitable for applying machine learning tools

 

Assignment

The page Assignment M1A1 Hands-on supervised learning gives a description of the assignment that we propose to fulfil these objectives. An alternative, as described above, is to define your own problem and work on that.

 

Assignment 2: Hands-on deep learning for NLP

Objectives

  • Hands-on experience with a powerful deep learning framework such as Keras/Tensorflow, Pytorch
  • Understanding of the building blocks in deep neural networks such as different type of neurons and layers, cost functions, initialisation. regularisation, back-propagation
  • Understanding of the influence of the most important hyper parameters and the size of the training data
  • Processed sequential real world data with a deep neural network
  • Made use of constructs such as RNN and LSTM

 

Assignment

The page  Assignment M1A2 Hands-on deep learning for NLP gives a description of the assignment that we propose to fulfil these objectives. An alternative, as described above, is to define your own problem and work on that.

 

Assignment 3: Exploration of the area

Objectives 

  • Broad knowledge of 
    • machine learning
    • knowledge representation and reasoning

Assignment

This assignment is mandatory for all and is described in the page Assignment M1A3 Exploration of additional topics. Unless impossible due to too few students taking the course in parallell (across all universities) this assignment must be carried out in groups.