Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
Next revisionBoth sides next revision
team:daniel_nyga [2017/04/13 13:35] – [Teaching] nygateam:daniel_nyga [2017/06/29 08:20] – [Dr.rer.nat. Daniel Nyga] nyga
Line 1: Line 1:
 ~~NOTOC~~ ~~NOTOC~~
-=====Daniel Nyga, M.Sc. (TUM)======+=====Dr.rer.nat. Daniel Nyga======
 | {{:wiki:daniel_nyga.jpg?0x180}} |||| | {{:wiki:daniel_nyga.jpg?0x180}} ||||
-|::: ||Research Staff\\ \\ ||+|::: ||Research Associate\\ \\ ||
 |:::|Room: |1.77| |:::|Room: |1.77|
-|:::|Tel: |--49 -421 218 64008|+|:::|Tel: |--49 -421 218 64010|
 |:::|Fax: |--49 -421 218 64047| |:::|Fax: |--49 -421 218 64047|
 |:::|Mail: |<cryptmail>nyga@cs.uni-bremen.de</cryptmail>| |:::|Mail: |<cryptmail>nyga@cs.uni-bremen.de</cryptmail>|
Line 10: Line 10:
  
 ====About==== ====About====
-Before I joined the Institute for Artificial Intelligence, I studied Computer Science at Technische Universität München, where I received my Master's degree in 2010 (with distinction). In February 2011 I started my PhD supervised by Prof. Michael Beetz at the //Intelligent Autonomous Systems// group at TUM, which I am now continuing at the IAI, University of Bremen.+Daniel Nyga is a postdoctoral researcher at the Institute for Artificial Intelligence (IAI), University of Bremen. Before he joined the IAI Bremen, he studied computer science at the Technical University of Munich, where he received a Bachelor's degree in 2008 and a Master's degree in computer science  in 2010. In 2011, he started his PhD supervised by Prof. Michael Beetz at the //Intelligent Autonomous Systems// group at TUM, which he has finished at the Institute for Artificial Intelligence Bremen with his thesis on the [[http://nbn-resolving.de/urn:nbn:de:gbv:46-00105882-13|Interpretation of Natural-language Robot Instructions: Probabilistic Knowledge Representation, Learning, and Reasoning]] (see below).
  
-I'm working on the import of action-specific knowledge from the World Wide Web into the knowledge bases of our mobile robots. In particular, my current research focuses on understanding natural language, in order to enable a robot to autonomously acquire new high-level skills by querying web pages such as eHow.com or wikiHow.com.+====Dissertation==== 
 +[[http://nbn-resolving.de/urn:nbn:de:gbv:46-00105882-13|{{:team:cover.png?200 |}}]]//Abstract//-- A robot that can be simply told in natural language what to do -- this  
 +has been one of the ultimate long-standing goals in both Artificial  
 +Intelligence and Robotics research. In near-future applications,  
 +robotic assistants and companions  will have to understand and perform  
 +commands such as "set the table for dinner", "make pancakes for  
 +breakfast", or "cut the pizza into 8 pieces." Although such  
 +instructions are only vaguely formulated, complex sequences of  
 +sophisticated and accurate manipulation activities need to be carried  
 +out in order to accomplish the respective tasks. The acquisition of  
 +knowledge about how to perform these activities from huge collections  
 +of natural-language instructions from the Internet has garnered a lot  
 +of attention within the last decade. However, natural language is  
 +typically massively unspecific, incomplete, ambiguous and vague and  
 +thus requires powerful means for interpretation. 
 +This work presents PRAC -- Probabilistic Action Cores -- an  
 +interpreter for natural-language instructions which is able to resolve  
 +vagueness and ambiguity in natural language and infer missing  
 +information pieces that are required to render an instruction  
 +executable by a robot. To this end, PRAC formulates the problem of  
 +instruction interpretation as a reasoning problem in first-order  
 +probabilistic knowledge bases. In particular, the system uses Markov  
 +logic networks as a carrier formalism for encoding uncertain knowledge.  
 +A novel framework for reasoning about unmodeled symbolic concepts is  
 +introduced, which incorporates ontological knowledge from taxonomies  
 +and exploits semantically similar relational structures in a domain of  
 +discourse. The resulting reasoning framework thus enables more compact  
 +representations of knowledge and exhibits strong generalization  
 +performance when being learnt from very sparse data. Furthermore, a  
 +novel approach for completing directives is presented, which applies  
 +semantic analogical reasoning to transfer knowledge collected from  
 +thousands of natural-language instruction sheets to new situations. In  
 +addition, a cohesive processing pipeline is described that transforms  
 +vague and incomplete task formulations into sequences of formally  
 +specified robot plans. The system is connected to a plan executive that  
 +is able to execute the computed plans in a simulator. Experiments  
 +conducted in a publicly accessible, browser-based web interface  
 +showcase that PRAC is capable of closing the loop from  
 +natural-language instructions to their execution by a robot.
  
-My work aims at building up action-specific knowledge bases from various knowledge sources, such as natural language, interactive computer games, observations of humans performing everyday activity or experience data of a robot.+====Projects==== 
 +Daniel Nyga's research interests revolve around topics on Artificial Intelligence and Data Science in general, as well as Machine Learning, Data Mining and Pattern Recognition techniques. In particular, he is interested in probabilistic graphical and relational knowledge representation, learning and inference methods, and in applications thereof in natural-language understanding, knowledge processing and robotics.
  
-{{research:actioncore.png?w=620&h=65&t=1357297411}}+He was involved in the European FP7 research projects [[http://www.robohow.org|RoboHow]] and [[http://www.acat-project.eu|ACAT]].
  
-Knowledge about actions and objects is represented as //Probabilistic Robot Action Cores (PRAC)//, which can be thought of generic event patterns that enable a robot to infer important information that is missing in an original natural-language instruction. PRAC models are represented in //Markov Logic Networks//, a powerful knowlegde represenation formalism combing first-order logic and probability theory. +He is the lead developer in the projects [[http://www.pracmln.org|pracmln]] and [[http://www.actioncores.org/|PRAC]].
- +
-I am involved in the European research projects [[http://www.youtube.com/watch?v=qQG3CkH27qc#t=118|RoboHow]] ([[http://www.robohow.org]]) and [[http://www.acat-project.eu|ACAT]]. +
- +
-I am also the lead developer in the projects [[http://www.pracmln.org|pracmln]] and [[http://www.actioncores.org/|PRAC]]. +
- +
-If you are interested in a student project in any of the above topics, please contact me via E-Mail or just drop into my office. +
  
 +His GitHub profile can be found [[http://www.github.com/danielnyga|here]].
  
  
Line 36: Line 70:
   * Data Mining/Knowledge Discovery   * Data Mining/Knowledge Discovery
   * Automated Learning/Understanding of WWW information   * Automated Learning/Understanding of WWW information
-  * Natural-Language Understanding+  * Natural-language understanding
  
 ====Teaching==== ====Teaching====
Line 64: Line 98:
 <author>nyga</author> <author>nyga</author>
 </bibtex> </bibtex>
 +
- Impressum - Datenschutz



Prof. Dr. hc. Michael Beetz PhD
Head of Institute

Contact via
Andrea Cowley
assistant to Prof. Beetz
ai-office@cs.uni-bremen.de

Discover our VRB for innovative and interactive research


Memberships and associations:


Social Media: