Během akademického roku jednou týdně přednáší zvaný přednášející
(z ciziny i od nás) o své vědecké práci.
Kolokvium, které probíhá na Fakultě informatiky, je otevřeno pro
vědeckou veřejnost. Termíny přednášek můžete najít v programu.
Úterý 14.00 - 15.00, D2, FI MU, Botanická 68a, Brno
25. 2. 2020
Ing. Vojtěch Mrázek, Ph.D., FIT VUT
Energy-efficient neural networks for embedded systems
Abstrakt:
Artificial neural networks are optimized for high-performance
computer systems. However, the inference path of the NNs is often
executed in small embedded systems such as special ASIC or FPGA
accelerators. Since these systems are typically battery-powered, the
energy consumption becomes crucial. In this context, this talk deals
with three topics. (i) Overview and challenges of hardware NN
accelerators. (ii) Error resiliency of neural networks. (iii)
Approximations of NN inference path for applications such as low power
image classifiers.
3. 3. 2020
prof. RNDr. Jiří Zlatuška, CSc., FI MU
Kurt Gödel as a Destruction & Revival Theme
Abstrakt:
Kurt Gödel's results can be viewed as a destruction of Formal
Mathematics as a discipline dealing with mathematical truths, and at the
same time as laying a foundation of subsequent revival of the field in
the form of Informatics as a discipline dealing with the nature of
computational side of the world (complementing the physical and
consciousness sides dealt with in other branches of scientific
disciplines). This talk is a slight extension of the talk given at the
opening of the "Tribute to Kurt Gödel 2020" international conference which
took place in Brno in January 2020.
10. 3. 2020
Piotr Micek Ph.D., Department of Theoretical Computer Science, Jagellonská univerzita, Krakov, Polsko
Planar graphs have bounded queue-number
Abstrakt:
Stacks and queues are fundamental data structures in computer science. But
what is more powerful, a stack or a queue? In 1992, Heath, Leighton, and
Rosenberg developed a graph-theoretic formulation of this question, where
they defined the graph parameters stack-number and queue-number which
respectively measure the power of stacks and queues to represent a given
graph. It is still open whether every graph has stack-number bounded by a
function of its queue-number, or whether every graph has queue-number
bounded by a function of its stack-number. Planar graphs were the simplest
class of graphs where it was unknown whether the queue-number is bounded.
(The stack number of planar graphs is at most 4.) We show that planar
graphs have queue-number at most 48. The key to the proof is a new
structural tool called layered partitions, and the result that every planar
graph has a vertex-partition and a layering, such that each part has a
bounded number of vertices in each layer, and the quotient graph has
bounded treewidth. We call the latter result the product structure theorem
for planar graphs. We will discuss further applications, including
adjacency labelings, nonrepetitive colorings, and p-centered colorings.
17. 3. 2020
dr. Renata Raidou, Faculty of Informatics, Technická univerzita ve Vídni, Rakousko
Visual Analytics in Cancer Treatment
Abstrakt:
My talk will focus on recently proposed solutions from the field of Visual Analytics for cancer treatment. One of the most common cancer treatment options is radiation therapy, which involves careful treatment planning through several complex processes and the incorporation of a multitude of patient data. As radiotherapy research tries to design more effective treatment strategies, it becomes essential for researchers to understand and to integrate all available patient- and tumor-related knowledge into the current workflow. My work focuses on increasing the – up to now, limited – insight and exploratory capabilities of radiotherapy researchers. In this talk, I will discuss comprehensive and comprehensible visualizations that contribute towards the interactive exploration, visual analysis and understanding of radiation therapy data and processes, creating a fertile ground for future research in radiotherapy treatment planning.
24. 3. 2020
dr. Nafaa Jabeur, Ph.D., Computer Science Department, German University of Technology in Oman, Halban, Omán
Collaborative Hazard Management Based on Intelligent Drones
Abstrakt:
Recent technological advances are enabling drones to collect crucial data
on events and objects of interest, particularly from areas where human operators have no
or limited access. However due to their limited flight times and reduced onboard
capabilities, increasing attention is being given to enabling fleets of drones to jointly
operate and achieve common goals which are beyond their individual capabilities. Within
this context, we are proposing a new approach were a group of drones intelligently
collaborate and self-adapt their actions to monitor ongoing hazardous situations.
31. 3. 2020
doc. Mgr. Zdeněk Dvořák, Ph.D., MFF UK
Approximation algorithms in classes with sublinear separators
Abstrakt:
Lipton and Tarjan proved that every n-vertex planar graph has a balanced
separator of order O(sqrt(n)) and gave a number of algorithmic applications of
this result. In particular, they showed that the maximum size of an
independent set in graphs with this property can be approximated arbitrarily
well in polynomial time. The idea of their algorithm applies to other
problems, but only subject to somewhat limiting restrictions (the optimal
solution must be of linear size and the problem must be defined by edge
constraints).
19. 5. 2020
doc. Razvan Beuran, Japan Advanced Institute of Science and Technology, Nomi, Japonsko
Research and Development on Cybersecurity Education and Training
Abstrakt:
Cybersecurity education and training via hands-on activities
are critical for increasing the resilience of government institutions,
organization and companies. In this talk I shall introduce the research
and development conducted in this area by the the Cyber Range
Organization and Design (CROND) NEC-endowed chair at JAIST. The main
focus will be on the integrated cybersecurity training framework, named
CyTrONE, that we designed and implemented with the goal of automating
the training content generation and environment setup tasks. In
addition, I shall discuss our activities related to security training
content development, as well as the IoT training system named
IoTrain-Sim. Both the systems and the content that we develop are
available as open source on GitHub and on the JAIST website.
Program kolokvií s abstrakty pro semestr Jaro 2020
In the last few years, several more sophisticated techniques for graphs with sublinear separators were developed; these techniques overcome the aforementioned restrictions and enable us to design approximation algorithms for much more general classes of problems. I will survey these developments and the outstanding challenges.