Assessment rules

• Two presentations relevant to machine learning or computer networks / optimization:

  • Block A – First presentation (20 min) – Overview of the assigned research topic + known methods for solving a given problem. Must be based on research papers including at least 3 published in the last 5 years

  • Block B – Second presentation (20 min) – Detailed overview of a significant research paper from the assigned research area (at least 140 points for AI/ML or at least 70 points for computer networks - https://www.gov.pl/attachment/7839ea33-4b51-4bfa-b642-fa347f315fda )

  • Paper sources:

    • www.googlescholar.com
    • Example journals for AI/ML:

    Knowledge-Based Systems

    Engineering Applications of Artificial Intelligence

    Information Fusion

    Neurocomputing

• Before each presentation, two persons are selected to ask a question (questions are graded)

  • plus — if the question is substantive and contributes to the discussion (2xplus = +0,5 grade)
  • neutral — questions that add nothing to the discussion but are in some way relevant to the topic
  • minus — if the question is not related to the presentation or the answer was stated during the presentation (2xminus = -0,5 grade)

• One unexcused absence is allowed (each subsequent one results in a 0.5 grade reduction). Failure to prepare the speech for the assigned date results in a reduction of the final grade by 0.5.

• Each presentation is graded based on its length, quality of prepared slides, quality of information provided and presentation skills. The final result is the average grade of both presentations.

• All presentation must be archived in the ePortal system before the seminar on which they are presented. Failure to comply with this requirement may result in the inability to present during the class or failing the class.

Topics

1. Anisotropic filtering of digital images,
2. Open-Set Recognition,
3. Dynamically imbalanced data stream classification,
4. Spectral Unmixing of Hyperspectral Remote Sensing Imagery,
5. Unsupervised and Agnostic Concept Drift Detection,
6. Multimodal data classification,
7. Classical and deep learning approaches to Natural Language Processing,
8. Machine learning under limited label access (semi-supervised learning, self-labeling, active learning),
9. Learning without label access (self-supervised learning),
10. Static and dynamic classifier selection,
11. Deep learning for tabular data,