Artificial intelligence tools

Master's seminars

Approximate reasoning 1000-1M00WA
Artificial intelligence and expert systems 1000-2N00SID
Big data mining and processing 1000-2M13DZD
Data mining 1000-2M03DM
Machine learning 1000-2N09SUS

Classroom

The seminar is devoted to the broadly understood subject of creating intelligent systems based on artificial intelligence methods and modern machine learning, in particular. Much of the discussion will focus on practical issues related to industrial and business applications of such systems, as well as the tools needed to keep these systems in constant operation. The description includes examples of such AI methods.

During the seminar, students present and discuss concepts from their own master's theses or present interesting AI-related topics, provided they have been approved by the seminar coordinators. There is also a list of proposed topics.

The seminar provides an opportunity to discuss any problems encountered during the writing of the MSc thesis, as well as potential directions for its future development.

The seminar is common to both computer science and mathematics.

During the seminar, students give presentations on selected topics related to AI methods/tools or directly on their master's thesis work. The instructors provide a list of proposed topics and offer consultative assistance. One meeting involves one or at most two presentations.

Example technical areas:

- neural network architectures, both the latest ones such as Transformer and KAN, and older ones such as LSTM and ResNet. This includes both general and specific models.

- transfer learning

- XAI techniques such as LIME, Shap, and Saliency Maps

- mathematical foundations of AI

- custering techniques, dimensionality reduction, and feature selection

- tree-based models

- bagging and boosting

- SVM models

- metaheuristics such as PSO and simulated annealing

- cvolutionary approaches such as evolutionary strategies and genetic algorithms

- knowledge representation methods

- methods for operating under uncertainty and incomplete information

- fuzzy logic and fuzzy reasoning

- issues related to ML Operations (MLOps), including maintaining, monitoring, testing, operationalizing, and deploying ML models to production

Example application areas:

- computer vision

- natural Language Processing (NLP)

- audio processing

- medicine

- biotechnology, biology, chemistry

- robotics

- finance

- games

- transportation problems, logistics

- autonomous vehicles

- combinatorial optimization

Modern scientific literature of the subject, including scientific journals (eg., Science, Applied Soft Computing, Information Sciences, IEEE Transactions on Neural Networks)and conference proceedings (eg., AAAI, IJCAI, CVPR, NeurIPS).

Due to the fact that the topics at the seminar can be new each year and because we discuss topics chosen by the students, the range of literature is extremely broad and dedicated to the specific selected topics.

More specific information is presented at the first meeting.

The students prepare and deliver seminar talks (K_U11) prepared on the basis of the newest publications concerning Natural Language Processing (K_U14, K_K08), among other also from conferences and journal of ACM and IEEE (K_K07). Many talks present interdisciplinary research, with prominent roles played by scholars from fields other than computer science (K_K02).

The student who presents a talk is expected not only to report on the paper, but also to express his/her own opinion on it (K_K06), those who listen are expected to participate in the discussion following the presentation (K_K02).

We also have a second kind of presentations, those related to the preparation of the Master's Thesis (K_U13). The first such talk is typically given soon after determining the Thesis' topic, and the student is expected to present a plan how he/she intends to gain the knowledge necessary to prepare the Thesis (K_K01, K_U15,K_K03).

1. Formal requirements: registering of a MS Thesis theme (1st year).

2. Delivering the presentations, at least one in each semester

3. At least 50% of attendance