CI-CS-IEEE logo.png

Alan Turing Seminars

Deep Learning: Introduction

by Rudolf Jakša
23.10. 2014
[ Slides ]

Introduction into Deep Learning techniques. The talk starts with the motivation - Vanishing Gradient Problem and Hinton 2006 paper with the Truck Hits House example; continuing with the RBM, autoencoders and denoising autoencoders. Also touching the convolutionary networks with Neocognitron and LeNet networks and the dropout technique. All topics are accompanied with links to original papers and comments to corresponding research in Košice.

Towards the Autonomous Hydroponic Systems

by Martin Paľa
12.6. 2014
[ Slides ]

This seminar will be focused on smart hydroponic and aeroponic systems. Very short introduction to both of these techniques in general will be given in the first part. We will specify the main problems and challanges at current stage of hydroponic system automation processes. The second part of the talk is aimed to introduce G.E.C.K project. The purpose of G.E.C.K project is to design and develop various hydroponic and aeroponic systems with high level of autonomy. During the talk we will present the main ideas of this project and provide short report on ongoing works. The ultimate aim of the seminar is to put the experts from fields of Botany, AI, Cybernetics and any other profession together to discuss the possible cooperation.

In addition, our first Flood and Drain hydroponic prototype will be uncovered. By the time you can find more information on the following link: or our facebook page.

Microsoft Cloud Services for KKUI

by Zuzana Krnáčová
22.5. 2014

Windows 8.1 and Office 365 presentation is suitable for every researcher and employer of the Technical University of Kosice. The main content of the presentation is introducing operating system Windows 8.1 and showing features that can simplify user experience. Introduction of Windows 8.1 is suitable for end users.

Second part of the presentation is Office 365. Office 365 is well known Office with extension to cloud services. Office 365 users can use file share based in cloud services provided by Microsoft. Part of the presentation will be introduction of Office 365 components.

After session of presentation for end users we will start with designing architecture for users on Department of artificial intelligence. This session is suitable for IT professionals and employees who will be involved in process of deployment.

Presentation will be streamed over Lync technology on link: Users with Lync client installed will be connected automatically. Users without Lync client will connect in web browser (ActiveX is required).

Machine Empathy: Towards Artificial Emotional Intelligence & Active Personalization in Social Human-Robot Interaction

by Mária Virčíková
15.4. 2014
[ Slides ]

Traditionally, a fundamental dream of artificial intelligence has been to create systems designed to emulate or simulate human behavior. Human intelligence expressed in both verbal and non--verbal communication has affect and empathy is as necessary to intelligence as pure reason. Could we create empathy for robots? And if, how it should be & where it would be useful?

Social robotics is an emerging part of service robotics but social robots are explicitly developed to support human--like interaction. The application potential is huge: robots for elderly, rehabilitation robots used in various processes of cognitive therapy, robots used in education or entertainment robots. However, the future goal is to move from the specific environments to create universal social companions. Current studies of social human-robot interaction use the 'Wizard of Oz' technique, where the WoZ refers to a person remotely operating a robot. If autonomous robots (operating without a person controling its behavior) are being used, their behavior is limited, usually preprogrammed for a set of tasks in specific scenarios.

During my PhD. I designed a methodology to create a so-called "affective loop," a system for social robot that simulates empathy - to cooperate with human more efficiently, to help him better,to motivate him in doing something. To be, from the user's point of view, empathic--like real social companions.

Obviously, I do not doubt the interdisciplinary character of this goal, where psychology plays a crucial role. However, one need not be a psychologist to know that there is a large personality difference in people´s expectations and preferences, ranging from environments & from person to person. My work gives a solution: PERSONALIZATION. Robots learn from humans how they what robot to behave. The affective loop consists of multimodal perception of user's social cues (especially body-based expressions), internal model to manage the artificial emotional state of the robot, planning the appropriate behavior and emotional expressions of the robot. All parts of the loop have the ability of personalization.

I argue that my work is a small step towards a big goal, which is the continuation of Gates idea: a SOCIAL ROBOT IN EVERY HOME. A social robot in the way that every user wants it to be.

The key research contribution of my PhD. work that will be exlained during the presentation is:

Design and implementation of a new concept of affective loop able of personalization during the interaction with users where the computational intelligence appears in form of: ARTMAP neural networks for the classification of users gestures into emotion categories, fuzzy logic for creation of metalogic of the internal emotional model of robots, principles of learning from demonstration to map non-verbal expressions from human to robot using Kinect and optimization of the expressions using the interactive evolutionary computation. The work also designs quantitative measurement of the effectiveness of human-robot cognitive cooperation.

Sentiment Analysis in Human-Robot Interaction

by Martina Tarhaničová
21.3. 2014
[ Slides ]

The main goal of the presentation is to overview the current state of the art of sentiment analysis and emotion detection. It describes lexical, corpus based and machine learning methods for detecting emotions in text. To distinguish between emotions, it defines emotional models deciding those emotions to include in Today there are three primary models used: categorical, dimensional and appraisal-based. Each of them is described and founded by research papers. The novelty of this work is proposing emotion detection from text in conjunction with robotics. The proposed system for emotion detection will be available as a Software as a Service in cloud computing, offering to combine all kind of problems - beginning by connecting the robot to the cloud until the right emotion from text is identified. The proposed system will become a small part of a larger system that falls within the field of human-robot interaction.

Use of Fuzzy Cognitive Maps in Robotics

by Ján Vaščák
12.2. 2014
[ Slides ]

Fuzzy cognitive maps have become thanks their comprehensive knowledge representation and a simple inference principle a considerably favourite means for knowledge experts in the last decade. Because of their properties enabling to capture complex chains of relations, including closed loops, this means finds the use also in the area of control and decision making in general and particularly robotics, too.

This presentation together with a brief theoretical description of fuzzy cognitive maps and analysis of a robot from the informatics viewpoint deals with some selected design examples of their use on individual decision levels of a robotic information system and hereby it offers a basic overview about possibilities and ways of their use in robotics tasks.

Cloud, Robotics, Object Recognition

by Daniel Lorenčík
[ Slides ]

In my talk, I will start with attempt to answer the question "What is the Cloud" as I think there is no consensus on this matter. I will continue with the pros and cons of the Cloud and how it can change our perception of the IT and computing in general. As this is the core of the motivation of my research and work here in CIT, I will continue with motivation for my research, what I want to achieve and why I consider it important. I will also include the past experiences with the cloud, my brief stay at Microsoft Slovakia, what have I already done, and what are my plans to the immediate and more distant future. I hope that my talk will stimulate discussion about clouds, robotics and the next course of the evolution of methods of AI.

Towards the Autonomous Food Production Systems

by Martin Paľa
[ Slides ]

The main motivation of this seminar is to provide overview information about progressive techniques and methods for producing green food with consideration for environmental factors and energy efficiency. The main idea behind the (semi) autonomous food production system is full automation, scalability, anytime-anyplace access monitoring and fault diagnostics for home or enterprise farming. The first part of this seminar will be dedicated to brief literature preview about hydroponics and aeroponics methods of growing plants, its common and distinctive features and the description of the needs for its automation. In the second part we will introduce general hydroponics/aeroponics control system architecture proposal (+ hydroponic prototype details) and its main features as power optimization, remote control or fault diagnostics.

Opinion Mining within Web Discussions

by Kristína Machová
[ Slides ]

The talk will introduce various problems, which can be solved by mining conversational content provided by "point-to-point" or "multicast" on-line web services (e.g., chats, web discussion forums, and IRC (Internet Relay Chat) commentaries, blog and micro-blog platforms, etc.). The focus will be on two problems: opinion classification and authority identification. Various approaches to processing of negations and intensifiers within the opinion analysis field will be briefly described. In addition, the novelty methods which have been designed on the basis of dynamic analysis and cyclic usage of n-grams will be presented. The implementations of these methods for the Slovak language will be mentioned.

Intelligent Cloud Robotics - Basic Concept

by Peter Sinčák
[ Slides ]

Talk will be focused to basic principles of Cloud Robotics which were set up by James Kuffner from CMU, Google Research and try to discribe conditions, technical environment and potential technologies used in Cloud Robotics. Intelligent Cloud Robotics goes for Learnable Cloud robotics which incorporate intra Cloud learning ability and also inter Cloud Learning ability from other Clouds or environment. Extrapolation to Brain bahaviour Cloud Robotics will be wrapping up the talk including impact to industry, technologies and core research methods.

Unification, Accessibility and Diagnostics of Devices

by Peter Szabó and Zlatko Fedor
[ Slides ]

We will talk about our upcoming software system telescope dealing with the problems of unification, accessibility and diagnostics primarily for the robot Nao and robotic drones platforms. There is a huge number of devices over the world which includes mobile devices, robotic systems, various sensors, kitchen appliance, etc... This devices creates heterogeneous environment with different user interfaces together with various programming tools needed to operate them. Unification, accessibility and diagnostic of such a devices is a solution which we will discuss in our seminar.


Older seminars can be found at the Seminar Archive page

Center for Intelligent Technology is equipped with 16 NAO Humanoid Robots for research and educational purposes towards Intelligent technologies.

Annually the branch of AI finishaprx. 25 people in Bc. and MSc. . PhD level is also very active - we do have 11 alumni in AI past 15 years in branch of AI.