Applied Image and Signal Processing
Analytics and Knowledge Discovery
| Semester | 1 |
|---|---|
| Academic year | 1 |
| Course code | AISM1AKDIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 2 |
| ECTS Credits | 3 |
| Examination character | immanent |
Lecture content:
Analytics, EDA Parallel Lines, Boxplots, Kernel Density Estimators, Basic Coding, Curse of Dimensionality, PCA, tSNE, Kmeans, hierarchical clustering, Spectral clustering, Distances and similarities
Learning Outcomes:
The module Analytics and Knowledge Discovery leads alumni to classical approaches on Exploratory Data Analysis for data with different kind of representation (numerical, categorical, text). For implementing a knowledge discovery process, they apply methods to reduce the dimension-ality of data, cluster it and apply various visualization methods. The course concentrates on unsupervised methodology.
Superior module:
Data Science & Analytics
Module description:
xxx
Data Science
| Semester | 1 |
|---|---|
| Academic year | 1 |
| Course code | AISM1DSCIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 3 |
| ECTS Credits | 5 |
| Examination character | immanent |
Lecture content:
Definition of Terminology in Data Science and Artificial Intelligence, Design Cycle, Extended Design Cycle, Sampling, Pre-processing, Normalization, Performance Measures, Cross Validation, Training Policies, K-nearest Neighbour and Minimum Distance Classifier, NLP Pre-processing and Features, Low Level Image Features
Learning Outcomes:
Upon completion of this course, alumni know about types and ingredients of data science projects, entitle their structure and identify different types of team members. They understand the concepts of data, models and algorithms and use specific language to describe data. They discuss the appropriateness of a data collection or intended data acquisition process with respect to a data science or artificial intelligence project. Alumni are introduced to the classical approach for extracting information from data with different kind of representation (numerical, categorial, one-hot or text). They collect, pre-process and visualize this data to gain basic data understand-ing. They follow the design cycle for supervised methodology by implementing data-specific feature generation, sampling of training and testing data, training selected (simple) classifiers and evaluating their performance. The alumni use state-of-the-art development tools and scalable technology and argue their approach content-wise.
Superior module:
Data Science & Analytics
Module description:
xxx
Digital Signal Processing 1
| Semester | 1 |
|---|---|
| Academic year | 1 |
| Course code | AISM1DSPIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 3 |
| ECTS Credits | 5 |
| Examination character | immanent |
Lecture content:
Theory of discrete signals and systems, discrete Fourier transformation, FFT, power density spectrum, discrete convolution and correlation, interpolation, calculations in z-domain, z-transfer func-tion, stability and frequency response of discrete systems, discretization of continuous systems (bilinear transformation, impulse invariant transformation), digital filters, principle and design of FIR filters, principle and design of IIR filters, IIR filter structures, quantization problems frequency transformations, simulation of signal processing algorithms and implementation of discrete sys-tems in lab environment (e.g. Matlab, Python, C)
Learning Outcomes:
Alumni understand the basic mathematical concepts to describe continuous and discrete time signals and systems and know the relations between time and frequency domain. They are famil-iar with the foundations of signal sampling and discretization and can apply important transfor-mations, e.g. Fourier-, Laplace and z-transformation. They understand basic algorithms in digital signal processing like FFT, convolution and correlation. They can transform continuous to discrete time systems e. g. with help of the impulse invariant or bilinear transformation and understand the restrictions. They have profound knowledge in designing and implementing digital filters and are also familiar with their applications. Alumni also have experience in simulation of DSP algorithms in a lab environment and are able to implement discrete systems with help of simulation software and low-level programming languages.
Superior module:
Digital Signal Processing
Module description:
xxx
Mathematics & Modelling
| Semester | 1 |
|---|---|
| Academic year | 1 |
| Course code | AISM1MAMIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 4 |
| ECTS Credits | 5 |
| Examination character | immanent |
Lecture content:
Vector valued functions on n-dimensional domains, vector fields, scalar fields, partial derivatives, gradient operator, Jacobi and Hessian matrix, directional derivative, Taylor series in several variables, critical points, local minima, maxima and saddle points, convex optimization and applications. Integral calculus, Pre-Hilbert (inner-product) space, (orthonormal-) basis and basis transformation, Eigenvalues, Eigenvectors, matrix decompositions and applications (PCA).
Learning Outcomes:
Alumni can apply functions in several variables to model problems. They are able to analyze the change behavior of these functions and to determine critical points. They can approximate complex functions by multidimensional polynomials (especially with tangent planes and second order Taylor polynomials). They are able to use gradient based methods to find local minima. They understand selected problems of convex optimization and can solve them with mathematical software. Alumni are able to calculate the most important matrix decompositions and apply eigenvalue theory to perform the principal components analysis for data. Alumni can solve multidimensional integrals. They understand the notion of a vector space (VS) with inner product and relate to it in different application areas. They master the coordinate transformation for the change of basis in finite dimensional VSs and are familiar with the relationship to Fourier analysis. They know selected application areas of the mentioned methods.
Superior module:
Mathematics & Modelling
Module description:
xxx
Applied Statistics
| Semester | 2 |
|---|---|
| Academic year | 1 |
| Course code | AISM2APSIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 3 |
| ECTS Credits | 4 |
| Examination character | immanent |
Lecture content:
Estimation theory: Point and interval estimators, maximum likelihood method, method of moments, parametric and non-parametric models (kernel density estimators, normal distributions, mixed models), statistical tests, study design and analysis of variance. Data visualization. Outlook: Random numbers and randomization; Graphical models and applications.
Learning Outcomes:
Alumni can apply methods of inferential statistics to data and communicate the results obtained both verbally and graphically. They can describe data with models and are able to represent dependencies of random variables with graphical models. They know statistical standards and are able to plan, conduct and document experiments. They know applications of random number generators in the area of generative models and can produce corresponding data with mathematical software.
Superior module:
Mathematics & Modelling
Module description:
xxx
Digital Signal Processing 2
| Semester | 2 |
|---|---|
| Academic year | 1 |
| Course code | AISM2DSPIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 3 |
| ECTS Credits | 5 |
| Examination character | immanent |
Lecture content:
Designing IIR filters with 2nd order sections, notch filters and comb filters with simulation tools (e.g. Matlab) and with low level programming language (e.g. C), principle and theory of adaptive FIR filters (LMS-filter) including implementation in low level programming language, quality enhancement with help of oversampling, polyphase filters, theory and simulation of sigma delta con-verter, numerical programming, fixed-point and floating-point number representation, floating point arithmetics, rounding, numerical analysis, basics of 2D signal processing.
Learning Outcomes:
Alumni know details in digital filter design such as advantages and disadvantages of different filter types and design methods. They understand the problem of quantization of filter coefficients and how to design 2nd order sections IIR filters. They know how to design special filters like notch, comb or median filters and are able to implement them in a low-level programming language (e.g. C). Alumni understand the concept of adaptive signal processing and can implement an adap-tive LMS filter e.g. for noise cancellation. In general, they can solve complex signal processing problems on a given hardware platform. Alumni understand the problems of numerical program-ming. They know common number formats and understand details of fixed point and floating-point arithmetic. They understand the principle of applying standard DSP algorithms also for 2D-signals (images).
Superior module:
Digital Signal Processing
Module description:
xxx
Machine Learning
| Semester | 2 |
|---|---|
| Academic year | 1 |
| Course code | AISM2MLGIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 3 |
| ECTS Credits | 5 |
| Examination character | immanent |
Lecture content:
Statistical Learning Theory, no free lunch, learning curve, loss functions, bias and variance; Models: Maximum Entropy (Logistic Regression), Artificial Neural Networks, SVM (Kernel SVM, Multi-Class SVM, One-Class SVM), Naive Bayes, Minimum Risk, AI-Applications
Learning Outcomes:
Alumni understand the assumptions and restrictions implied by a specific model choice in view of statistical learning theory setup and the "no free lunch" theorem. They select from a collection of well-known and widely available ML algorithms, accordingly, parameterize models and assess the impact of different design choices on the network complexity of neural networks. Alumni detect overfitting and underfitting during the training process and take corresponding counter measures such as regularization. They apply the machine learning models to different types of data (text, images, numerical) for tasks such as classification, representation learning and object detection and thereby construct examples of AI (artificial intelligence) systems.
Superior module:
Data Science & Analytics
Module description:
xxx
Agile Project Management
| Semester | 3 |
|---|---|
| Academic year | 2 |
| Course code | AISM3APMIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 2 |
| ECTS Credits | 3 |
| Examination character | immanent |
Lecture content:
The focus is on the creation of software engineering projects to cope with the digitalization of companies. Project management and software engineering skills are to be applied in the practical implementation. Among other things, business case & product innovation (using business canvas & value proposition canvas), project organization (process-oriented and agile procedure models, roles, work packages, milestones, reporting, results). The project implementation is carried out with templates from Software Engineering for the development, documentation and communication of software architectures using ARC42 (Context, Requirements, Constraints, Concept of Operations, Major building blocks/components, Block diagram, interfaces, workflow, control flow).
Learning Outcomes:
Alumni can apply theoretical and practical project management and software engineering skills in a team, based on the practical implementation of a continuous software engineering project.
Superior module:
Applied Sciences and Methods
Module description:
xxx
Ethics & Sustainability
| Semester | 3 |
|---|---|
| Academic year | 2 |
| Course code | AISM3ESAIL |
| Type | IL |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 1 |
| ECTS Credits | 1 |
| Examination character | immanent |
Lecture content:
The need for professional-ethical orientation has never been as great as it has become in the past decade. At this stage, we are being confronted with the topic of ethics from all directions: bioethics, medical ethics, animal ethics, ethics and politics, ethics and economy, ethics as a school subject instead of religion ....from personalized ethics to environmental ethics, from day-to-day to systems ethics.... our very existence seems to be sailing in a sea of ethical and morally charged issues - particularly because the two terms - ethics and sustainability - are being used more and more ambiguously and prolifically. This symposium will therefore attempt to shed some light on the question of terminology and to sensitize participants to the questions behind professional ethics and sustainability.
Learning Outcomes:
After successfully completing the symposium, alumni are able to analyse and reflect on ethical-moral dilemmas; to evaluate opinions from a lecture in their own context of action; to argue social issues with a view to their own professional environment; to articulate and justify their own opinion in the group discussion.
Superior module:
Applied Sciences and Methods
Module description:
xxx
EC: Applied Natural Language Processing
| Semester | 3 |
|---|---|
| Academic year | 2 |
| Course code | AISM3ANLIL |
| Type | IL |
| Kind | Elective |
| Language of instruction | English |
| SWS | 1 |
| ECTS Credits | 2 |
| Examination character | immanent |
Lecture content:
Methods: Dialog-based Agents and Systems; Artificial Intelligence; task-oriented dialog systems and chatbots; Natural Language Generation, Interaction and Understanding; Question Answering, Slot Filling. Applications: Dialog systems and chatbots. Tools: Python, scikit-learn, nltk, ten-sorflow/keras/PyTorch, dialogflow.
Learning Outcomes:
Alumni are aware of the difference between task-oriented systems and dialog systems. They develop algorithms for generating natural language targeted at different tasks (slot filling, question answering) or for conversational purposes. Alumni know about existing tools for the development of dialog systems, their differences and how to integrate these tools into other applications such as social media.
Superior module:
Selected Topics in Applied Image and Signal Processing
Module description:
xxx
EC: Applied Reinforcement Learning
| Semester | 3 |
|---|---|
| Academic year | 2 |
| Course code | AISM3ARLIL |
| Type | IL |
| Kind | Elective |
| Language of instruction | English |
| SWS | 1 |
| ECTS Credits | 2 |
| Examination character | immanent |
Lecture content:
Deep RL, Reinforcement Learning Algorithms, Model-based RL, RL by use of Physics Engines
Learning Outcomes:
Alumni identify problems for model-base and model-free reinforcement learning, apply suitable algorithms and assemble solutions using toolboxes. They know how to use real-life simulations by physics engines for Reinforcement Learning and know the challenges when switching to robots or other hardware. They discuss current trends and upcoming areas of application.
Superior module:
Selected Topics in Applied Image and Signal Processing
Module description:
xxx
EC: Natural Language Processing
| Semester | 3 |
|---|---|
| Academic year | 2 |
| Course code | AISM3NLPIL |
| Type | IL |
| Kind | Elective |
| Language of instruction | English |
| SWS | 2 |
| ECTS Credits | 3 |
| Examination character | immanent |
Lecture content:
Methods: Natural Language Processing with Deep Neural Networks, e.g. Recurrent Neuronal Networks, Attention-Models, Transformers or BERT. Contextualized representations, Subword tokenization, Beam Search.
Learning Outcomes:
Alumni apply attention-based models for natural language processing and implement appropriate networks for applications in areas such as machine translation and sentiment analysis in social networks. Building on previously acquired skills in pre-processing text data, they use contextualized text representations and complex network architectures. They are able to decide on network parameters and design appropriate for the problem at hand and know the limits and areas of application of the respective algorithms.
Superior module:
Selected Topics in Applied Image and Signal Processing
Module description:
xxx
EC: Reinforcement Learning
| Semester | 3 |
|---|---|
| Academic year | 2 |
| Course code | AISM3RILIL |
| Type | IL |
| Kind | Elective |
| Language of instruction | English |
| SWS | 2 |
| ECTS Credits | 3 |
| Examination character | immanent |
Lecture content:
Markov Decision Process, Definition of RL, Components of RL (Agent, Policy, Model), Model and Non-model based RL, Optimization of RL, Deep RL, Reinforcement Learning Algorithms
Learning Outcomes:
Alumni identify problems suited for reinforcement learning, find suitable models and assemble solutions using toolboxes. They distinguish and differentiate between different setups based on input data type and assumptions on the environment and select corresponding algorithms and metrics. Using Deep Learning methodologies, the alumni design, optimize and evaluate deep reinforcement learning for a set of classical problems. They discuss current trends and upcoming areas of application.
Superior module:
Selected Topics in Applied Image and Signal Processing
Module description:
xxx
Master Exam
| Semester | 4 |
|---|---|
| Academic year | 2 |
| Course code | AISM4MAEDP |
| Type | DP |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 0 |
| ECTS Credits | 2 |
| Examination character | final |
Lecture content:
Defensio, Technical examination talks
Learning Outcomes:
The alumni are able to present and discursively defend the hypotheses and solution approaches developed in the master thesis. They are able to establish cross-references to contents of the study program.
Superior module:
Master Thesis & Master Exam
Module description:
xxx
Master Seminar 2
| Semester | 4 |
|---|---|
| Academic year | 2 |
| Course code | AISM4MASSE |
| Type | SE |
| Kind | Compulsory |
| Language of instruction | English |
| SWS | 1 |
| ECTS Credits | 2 |
| Examination character | immanent |
Lecture content:
Discursive defence of parts of the master thesis in group situations; presentation of scientific work as part of the state-of-the-art discussion for the thesis¿ topics; discussion of recent research re-sults in connection with colleagues' theses.
Learning Outcomes:
The alumni are able to present and discuss their own scientific work in a peer group situation. They can argue logically and in line with scientific standards as well as understand the importance of a methodical approach.
Superior module:
Master Thesis & Master Exam
Module description:
xxx
Master Thesis
| Semester | 4 |
|---|---|
| Academic year | 2 |
| Course code | AISM4MATIT |
| Type | IT |
| Kind | Diploma/master thesis |
| Language of instruction | English |
| SWS | 0 |
| ECTS Credits | 23 |
| Examination character | immanent |
Lecture content:
Developing and elaborating on the research questions and establishing a contentwise argumentation of a topic in applied image and signal processing with special consideration of a scientifically sound and structured presentation reflecting the current state of the literature.
Learning Outcomes:
The alumni are able to independently write sound academic papers based on common international standards. They can proceed methodically and systematically. They can analyse and present problems, provide solutions as well as formulate these appropriately and critically scrutinise them. The alumni are able to defend their approach.
Superior module:
Master Thesis & Master Exam
Module description:
xxx
| Legend | |
| Semester | Semesters 1, 3, 5: courses held only in winter semester (mid-September to end of January), Semesters 2, 4, 6: courses held only in summer semester (mid-February to end of June) |
| SWS | weekly contact hours over 14 weeks in semester (example SWS 2 equals 28 contact hours for the whole course |
| ECTS Credits | Work load in ECTS credits, 1 ECTS credit equals an estimated 25 hours of work for the student |
| Type | BP = Bachelor final exam DP/MP = Master final exam IL = Lecture with integrated project work IT = Individual training/phases LB = Lab (session) PS = Pro-seminar PT = Project RC = Course with integrated reflective practice RE = Revision course SE = Seminar TU = Tutorial UB = Practice session/Subject practical sessions VO = Lecture |