Logo INSO TARO Icon close

TARO (Taming Robots)
research in the area of artificial intelligence and machine learning

Main Research Areas

  • Applied Artificial Intelligence
  • Optimization/Automation with AI Technologies
  • Trust & Transparency
  • Social Responsibility
  • Privacy & Data Protection
  • Reproducibility & Accountability

Topics for Projects / Bachelor / Diploma Theses

Bachelor Thesis, Computer Science Project: Machine Learning Platform with Docker and Kubernetes

Design an implement a state-of-the-art machine learning platform using open-source tools.

  • Compare current state-of-the-art machine learning platforms and define a minimal set of required modules
  • Implement a pipeline using those modules with open-source tools
  • Examples Modules: Storage, Data Ingestion, Data Preparation, Model Training, Model Versioning, Model Deployment, Monitoring, Verification

Project: Play Arcade Games with Reinforced Learning

Arcade games form the 80s are fun and have very handy features: The graphics are simple, the controls are very limited and the game mechanic is in general very simple. This makes it perfect for a computer to teach itself playing a game.

Teach the computer to play a arcade game.

  • find a game suitable to be controlled by a computer
  • research on reinforcement learning for games
  • select framework (e.g. [1], [2], [3])
  • implement learning and training (e.g. [4])
  1. https://github.com/google/dopamine
  2. https://aws.amazon.com/about-aws/whats-new/2018/11/amazon-sagemaker-announces-support-for-reinforcement-learning/
  3. https://github.com/uber-research/atari-model-zoo
  4. https://deepmind.com/research/dqn/

Bachelor Thesis, Computer Science Project: Automated Processing of Paper Forms

Currently a lot of paper forms are submitted on a daily basis. This could either be as a hard copy by mail or scanned by e-mail. In any case, a manual processing of the form needs to happen which includes:

  • detecting the type of the form
  • extracting the data from the form
  • verify authenticity

The target of this thesis is to explore solutions to implement parts or all above step automatically. Various approaches should be explored including but not limited to traditional media understanding, machine learning and deep learning approaches.

Challenges:

  • OCR detection: Text can be written on a computer, but could also be hand written
  • Form detection: Forms can be in different languages and there might be variations of forms which differ only in single pages of a multipage form
  • Blob extraction: Pictures or drawings might be part of the form which need to be detected and extracted as a whole
  • Form versioning: Forms might change over time and there are overlapping times where both form versions need to be processed

Diploma thesis: Creating test data for ML algorithms

Testing ML and AI algorithms is not an easy task. Often we cannot simply use real data as they are covered by data protection laws. In such cases we need to create test data ourselves. Or we might try to anonymize existing data.

  • What are the important points to take care of to not introduce bias in the data?
  • How can we A/B test our results?
  • What are the limits and how big does the sample size have to be?
  • How do manually tinkered solutions compare to more advanced ones like Generative Adversarial Networks?
  • When does the later make sense?

The target of this thesis is to generate test data for an AI use case. This could for example be forms which are processed and categorised by another project. In this case we would need to create various real-looking samples of various documents. E.g. birth certificates, marriage certificates, etc. These will later get scanned in and categorised by another project.

Other Topics

  • Software Defined Storage im Cloud Zeitalter
  • Streaming bei Anwendungen von AI
  • Debugging und Verifikation von AI Output
  • Moralische Verantwortung/Ethics & Bias von AI
  • Vergleich von Search Engines
  • Continuous Delivery mit Machine Learning
  • ... or ... bring your own topic!

Office location: Wiedner Hauptstraße 76, Stiege 2, 2. Stock; 1040 Wien
Office hours: by appointment per e-mail
Fax: 01 / 5872198
E-Mail: taro@inso.tuwien.ac.at