Intendierte Lernergebnisse
Methods & GoalsThe aim of FOML is to communicate the fundamental principles and classicaltechniques of Machine Learning to students, wich is essential to understandmodern Deep Learning techniques.Besides theoretical input in form of slides and tutorials, students will work onbasic Machine Learning problems i.e. produce code in Python. Beyond creatingand training the model, students will learn how to conduct a ML study frombeginning to end, from problem definition to containerization using Docker.In view of tools, this course will make use of open-source software: Python,Scikit-Learn, Keras and possibly also PyTorch.
Lehrmethodik inkl. Einsatz von eLearning-Tools
Methods & GoalsThe aim of FOML is to communicate the fundamental principles and classicaltechniques of Machine Learning to students, wich is essential to understandmodern Deep Learning techniques.Besides theoretical input in form of slides and tutorials, students will work onbasic Machine Learning problems i.e. produce code in Python. Beyond creatingand training the model, students will learn how to conduct a ML study frombeginning to end, from problem definition to containerization using Docker.In view of tools, this course will make use of open-source software: Python,Scikit-Learn, Keras and possibly also PyTorch.
Inhalt/e
This course on fundamentals of machine learning - FOML - intends tocommunicate the basics of classical Machine Learning theory to Math, IT, andICT students. Building on basic Math skills students had gained in high school,FOML introduces students to the broad field, teaches basic principles andclassical ML techniques (SVM, Random Forest, etc.), which is essential tounderstand state-of-the-art Deep Learning techniques. As showcases, thiscourse will deal with everyday Machine Learning problems of the EnergyInformatics research field such as forecasting of time series, detectinganomalies in energy grids and clustering of consumers based on householdappliances.It is intended to have FOML laid out as conventional "Kurs" with 2 SWS and3ECTS, starting in Sommersemester 2023. Though the case studies will focuson energy data, this course welcomes students of any technical background.Modules1. Introduction to Machine Learningi. Basic Terms and Definitionsii. Supervised vs. Unsupervisediii. Bagging vs Boostingiv. Overfitting vs Underfitting2. Module on Regression and Classificationi. Linear Regression / Least Squaresii. Support Vector Regressioniii. Naive Bayesiv. k-Nearest Neighboursv. Decision Treevi. Ensemble Methodsvii. Random Forest3. Module on Clusteringi. Intro to unsupervised learningii. k-meansiii. Hierarchical clusteringiv. DBSCAN4. Module on Featuresi. Dimensionality Reductionii. Feature Engineeringiii. Model Selectioniv. Parameter tuningv. Pre/Postprocessing5. Module on Modern Technical Aspectsi. Containerization using Dockerii. Deploying ML models on the cloudiii. CI / CD life cycle6. Outlook on modern Machine Learningi. Intro to Neural Networksii. Brief Review of Deep Learning (LSTM, GRU, Transformer)iii. Review of recent breakthroughs