Machine Learning
2022/2023- Purpose and learning objectives
Machine Learning (ML) is a deeply fascinating topic. Here developers can write programs that "magically" solve difficult problems.
Knowledge
Feed some data to the algorithm, and a result will appear. If you are not satisfied, then add more data or train the model. Repeat.
In this course you will learn how to use Machine Learning (ML) to solve problems of different kind. We will start with a simple problem like how to train a model to behave like a logical AND operator. From there we move on to predict height based on age and to predict people's political views based on socio-economic data.
This course will focus on understanding the fundamentals of ML. We will solve the same problem in 3 different ways:
First using Google Sheets, then a Deep Neural Network Simulation Tool and finally using Python.
For the mandatory task, students will define their own project. It can be based on any of the following:
Deep Neural Network (DNN)
Reinforcement learning
Azure Machine Learning Studio
TensorFlow.js
Convolutional Neural Networks (CNN)
Recurrent Neural Networks (RNN)
The goal is for the student to get inspired and find their own problem to solve.The student knows how a computer can learn.
Skills
The student knows several ML methods, such as Regression, Classification, Clustering and Deep Learning. And hence the concepts of Supervised Learning and Unsupervised Learning.
The student knows tools such as Numpy and Pandas.The student can use ML to solve practical tasks.
Competences
The student can prepare a dataset from raw data. This includes Data Cleaning, Feature Selection, Data Transforms and Feature Engineering.
The student can make visual representation of data to select features. The student can train and test a model.
The student can use an ML-framework to solve a practical task.The student knows how to select a relevant ML method to solve a given problem.
The student can assess the quality and efficiency of a solution.
The student can find and use relevant online resources and documentation to build a ML solution. - Type of instruction
Class presentations of new material. In-class exercises. Group work and individual assignments.
- Subject/module requirement for
participation
Academic requirement for participation
The student must know programming at a level of 3rd. semester KEA Computer Science APEquipment needed to participate
Windows is required for the Simulation Tool. For everything else Win/macOS/Linux is OK. - Exam
The learning outcomes of the exam are identical with the learning outcomes of the subject(s)/modul(es)
Prerequisites for access to the examinationThere will be 2 mandatory exercises. They must be handed in and approved, for the student to gain access to any exam. The 2nd mandatory exercise may be used as an exam project.Exam in one or more subjectsSubject/module is tested standaloneType of examOral examinationOral 30 minutes exam (incl. greading) with internal censor.
○ 10 minutes presentation
○ 15 minutes Q&A
Student presents exam project:
○ Project demo
○ Explain interesting parts and choice of technologies
○ Q & A about project and curriculumIndividual exam or group examIndividualExam languagesEnglishDuration30 min.Type of evaluation7-point grading scaleExaminersInternal censure
In the subject Machine Learning you will receive 64 hours of instruction, which corresponds to 85 lessons (1 lesson = 45 min.) and 23% of your total workload for the subject.
The teaching primarily consists of the following activities: classroom teaching, exercises.
The preparation primarily consists of the following activities: exercises.
Read about KEAs Study Activity Model
*KEA can deviate from the number of hours if this is justified by special circumstances