Intelligent Systems

Computational Intelligence

teach

Lecture (infCI-01a)

4 SWS, ECTS Credits: 8
Language of instruction: german
Time: Monday 14:15-15:45 and Wednesday 10:15-11:45
Lecture period winter term 2021-2022: 18.10.2021 - 22.2.2022
For more information UnivIS or in Modulinformationsystem


Exercise (ÜinfCI-01a)

2 SWS
Language of instruction: german
Time: Wednesday 12:15-13:45
Lecture period winter term 2021-2022: 18.10.2021 - 22.2.2022
Fore more information UnivIS

 

Summary

The term "Computational Intelligence" (CI) describes a sub-area of ​​artificial intelligence. Essentially, it summarizes three biologically motivated areas of information processing: Based on algorithms of fuzzy logic and artificial neural networks as well as on evolutionary algorithms, the aim is to master complex systems and combine them with other, typically biologically-inspired, processes. Originally coined in the 1990s by the Institute of Electrical and Electronics Engineers (IEEE), the term is now often used synonymously with soft computing. All sub-areas included have in common that they make mechanisms of natural (i.e. in particular biological, physical or social) problem-solving strategies usable for mathematical or engineering-technical questions. The aim is not a direct transfer or "technical copy", but an understanding and imitation of the basic mechanisms. The methods developed in this way are in contrast to exact mathematical methods - one rather freely follows the motto: "What works is allowed".

 

Learning goals

The aim of the event is to provide an initial overview of the field of computational intelligence in theory and practice. Building on this, students should be given a basic understanding and appropriate approaches so that the following goals in particular can be achieved:

  • The students have a basic understanding of the complexity of technical systems and know how it can be mastered.
  • The students know how seemingly complex relationships can be easily described with mechanisms of computational intelligence.
  • Techniques from the field of Computational Intelligence / Soft Computing and their advantages and disadvantages in comparison are known.

 

Workload

60 hours of lectures, 30 hours of face-to-face exercises, 150 hours of self-study

Bachelor Seminar - Self-organised Systems

fish

Bachelor seminar (infBSemIS-01a)

2 SWS, ECTS studies: 4
language of instruction: german
Time and place: to be announced
For more Information UnivIS or in Modulinformationsystem

 

Summary

The seminar is dedicated to the topic of "Self-organised Systems". In the context of this seminar, this is understood to mean sets of autonomous and technical units that work together to achieve a common goal. Different questions arise that have to be mastered.

The seminar in the winter semester 2021/22 is dedicated to the consistent, fast and reliable storage and the associated finding of content within such self-organized systems. This is examined using algorithms from peer-to-peer computing.

The seminar in summer term 2022 is dedicated to the cooperation of autonomous intelligent systems and the associated interaction patterns within such self-organized systems. This is examined using examples from autonomic and organic computing.

 

Learning goals

The students learn to independently acquire complex issues using the compact presentation, as is customary in scientific publications (i.e. papers, articles or specialist books) and to prepare them in an understandable form.

 

Workload

30 hours of participation in the seminar, 90 hours of self-study

Seminar - Deep Learning

dl

Seminar (infSemDeLea-01a)

2 SWS, ECTS studies: 5
Suitable for ERASMUS / exchange students, language of instruction: English
Time and place: irregular rotation
Lecture period winter term 2021-22
More information see UnivIS or in Modulinformationssystem

 

The KickOff event for the Master Seminar "Deep Learning" will take place on Thursday, September 23, 2021 at 2 p.m. as a zoom call. There, the seminar schedule will be explained and topics will be assigned. In preparation for the topic assignment, the starting papers for the individual topics are available. If you are interested in the event, already want to reserve a place or have any questions, please contact Prof. Tomforde directly.

 

General information

In this seminar important work of the last years in the field of Deep Learning will be discussed. The seminar is offered in cooperation with the "Multimedia Information Processing" workgroup. The main focus is therefore on problems in the field of computer vision and active learning. For example, topics such as object detection, image description, inpainting, segmentation, but also novel learning methods are discussed.

Subject / Topics (Entries that are grayed out are already assigned):

 1. Object Detection / End-to-End Object Detection with Transformers (link)

 2. Object Detection / Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation (link)

 3. Image Generation / On the Spectral Bias of Neural Networks (link)

 4. Image Generation / PixelCNN (link)

 5. Disentanglement / Guided Variational Autoencoder for Disentanglement Learning (link)

 6. Online Learning / Towards continuous actions in continuous space and time using self-adaptive constructivism in neural XCSF (link)

 7. Online Learning / Off-Policy Deep Reinforcement Learning without Exploration (link)

 8. Baysian NN / Weight uncertainty in neural networks (link)

 9. Quantification Learning / Automatic plankton quantification using deep features (link)

10. 3D Reconstruction / Recurrent Neural Network for (Un-)supervised Learning of Monocular Video Visual Odometry and Depth (link)

11. Image Registration / SuperPoint: Self-Supervised Interest Point Detection and Description (link)

12. Semi-Supervised Learning / MixMatch: A Holistic Approach to Semi-Supervised Learning (link)

13. Semi-Supervised Learning / SCAN: Learning to Classify Images without Labels (link)

 

A previous successful participation in the course Inf-NNDL: Neural networks and deep learning is required.

The assessment in this module consists of the preparation of a written elaboration, the review of two works of your fellow students, as well as a presentation with subsequent discussion.

The complete course will be held in English. The elaboration, reviews and presentation (incl. discussion) will be done in English.

If you have any questions, please contact Johannes Brünger or Simon Reichhuber

 

Written elaboration

Your first task is to read and understand the paper assigned to you. You should also consider quoted literature that is necessary for understanding. Then you should summarize the contents of your paper in your own words. We expect about 12-20 pages. Again, you should not limit yourself to your paper, but also include accompanying literature.

The written elaboration is submitted via the EasyChair conference system.

 

Reviews

In a review you should assess whether or not the author has fulfilled the task of summarising the given paper. Of course, you also have to consider the original paper. The reviews are also carried out via the EasyChair conference system. You will be assigned two papers, which you should edit as follows:

Summary: Here you should briefly summarize the content of the elaboration. This shows that you as a reviewer have understood what it is all about.

Paper strength: Here you should list what you liked about the elaboration. So if you had to defend your colleague, these would be your arguments why the elaboration is good.

Paper weaknesses: Here you should list what you didn't like about the elaboration. So if you had to defend the rejection of the work, those would be your arguments why the elaboration is not good.

Preliminary evaluation: A summary of your opinion with arguments as to whether the task was fulfilled or not.

Overall evaluation: That summarizes again in a grade whether the elaboration has fulfilled the task or not. This recommendation will not be passed on to the author.

The reviews are part of the assessment in this seminar and are therefore included in your grade. We will evaluate how conscientiously you have reviewed the work of your fellow students.

 

Presentation

During the presentation, you should give a 20-minute presentation of the contents of the paper assigned to you. Afterwards we would like to have a short discussion about the contributions of the paper to Deep-Learning research. You as the speaker should lead this discussion and, if necessary, start it with questions you have prepared yourself. We plan to hold this mini-conference online.

 

Workload

30 h attendance, 120 h self-study

 

Master Project - Intelligent Systems

mp


Exercise (infMPInS-01a)

4 SWS, ECTS Credits: 10
Language of instruction: english
Time: Thursday 09:00 - 11:30 place to be announced
Lecture period winter term 2021-2022: 18.10.2021 - 22.2.2022
For more information UnivIS or Modulinformationssystem Informatik

 

October, 5th 2021 at 2:15 pm: Kick-off meeting for the Master Project Intelligent Systems  via this zoom link.
Meeting ID: 993 7924 7325 Passcode: 900778

 

Scope

This project is offered as a master project (If you are interested in a Bachelor/Master Thesis within the scope of this project, please contact Torben Schnuchel for more information)


Group

This project is offered as a group task; a team of 3 to 6 students is required (deliverables are adjusted based on the number of participants).

 
Motivation

Smartphones nowadays are equipped with a variety of sensors that developers have access to (Accelerometer, GPS, Bluetooth, ...). Since they collect this data from individuals, they are particularly well suited in crowdsourcing scenarios where multiple participants share data to build a central knowledge base. One exciting application scenario of crowdsensing is the monitoring of traffic to infer information about traffic flow, road conditions, crashes, etc. in a distributed manner.
 

Task

What you are working with:
- Android Framework
- Open Source IOT Backend
 

What is your task as a group?

In this scenario, your imaginary customer is a city administration that wants to create a solution to collect information about the street network of a city. They want to enable automatic detection of road damage and thus be able to coordinate and optimize road maintenance. They want to try an approach that is based on crowdsourcing; this way, people can choose to contribute their data to the city through an app while driving. The shared data consists of the raw readings of the smartphone's sensors instead of aggregated statistics so that analysts can extract precise patterns from the data.


1) Create a dataset

You will be provided with a client/server setup to record data consisting of a sensing framework for Android smartphones and an open-source IOT backend. You as a team will then plan a data recording by specifying a track with varying street conditions (asphalt, sand, potholes, ...) and finally create a dataset on that specific track. To train a model for detecting the condition of the road approx 30-60min of data per student should be sufficient.

2a) Implementation/Evaluation of Road Surface prediction

- Study scientific literature and identify 3 models (e.g. different architectures of neural networks) that you will train on the task of road surface detection
- Train the models on the task of predicting the road surface and adjust the hyperparameters to the task

2b) Implementation/Evaluation of Privacy

- This kind of data sharing comes with severe privacy issues. Instead of sharing the raw data, you will investigate, if it is possible to
    1) predict the user from the raw data
    2) predict the user from intermediate representations of the model (e.g. activation of hidden layers)

2c) Optional: Evaluation Pruning (depending on the number of students)

Motivation: Deploy road detection models on smartphones. To do so, the models must shrink in terms of the number of parameters to account for the limited computing resources of smartphones.
- Rank the parameters of the networks according to some measure (e.g. magnitude, activation, ...)
- Evaluate the performance of networks with varying degree of pruning


Fundamentals of the module

Besides the actual development of the software, the module aims at practical experiences in the following aspects:
- Specification and definition of “product” using standard software engineering tools
- Setup of scientific experiments using standard Data-Science libraries like numpy/pandas/keras/tensorflow/...
- (Self-)Organisation as a team, management of the process, deadline supervision
- Entire software development process until delivery to the customer
- Documentation of the product
Your supervisors are continuously monitoring the process in the role of a customer, i.e. you are expected to regularly demonstrate the progress


Registration

Interested students may contact the group (i.e., Torben Schnuchel) for more details.