CoSCo News

ICL research group will apply probabilistic machine learning to quantum mechanics based material design

troos — Thu, 26 Oct 2017 09:39:14 +0000

The Academy of Finland has granted funding for a 2-year project on "Probabilistic Machine Learning for Quantum Mechanics-Based Material Design". The funding is awarded to the MachQu consortium consisting of the Information, Complexity and Learning (ICL) research group lead by Assoc Prof Teemu Roos at HIIT and a team of researchers at the Department of Physics, University of Helsinki, lead by Dr Flyura Djurabekova and Prof Kai Nordlund.

The MachQu project will develop new probabilistic programming and active machine learning methods, and apply these to a currently very active topic in materials physics. It utilizes state-of-the art machine learning methods to develop new efficient quantum mechanics- based simulation tools of atomic migration processes in materials in the extreme conditions of fusion reactors and particle accelerators. The results of the materials simulations can be used for knowledge-based design of complex alloyed materials that can withstand extreme conditions.

Contact:

Teemu Roos, Associate Professor, Department of Computer Science, University of Helsinki

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Contact person:

Teemu Roos

Roos group teams up with researchers in Turku under an Academy funded 4-year project on Tensor-Based Machine Learning

troos — Wed, 21 Jun 2017 07:49:14 +0000

The Academy of Finland has granted funding for a 4-year project on "Tensor-Based Machine Learning for Big Data with Inherent Dependencies". The funding is awarded to the TensorML consortium consisting of the research groups of Assistant Professor Tapio Pahikkala at the University of Turku and Associate Professor Teemu Roos at HIIT.

Machine learning (ML) has recently achieved a lot in areas where the standard assumptions about the data hold and the amount of training data available is large. However, it still faces many challenges in areas where we would need it the most. The commonly made independently and identically distributed (IID) assumption about data is rarely holds in practice, violating the guarantees about ML methods relying on the assumption. In this project, novel tensor-based ML methods turn the IID violations into our advantage so that both sample complexity and computational complexity are decreased, and the reliability of the prediction performance estimates are improved. Immediate applications include zero-shot learning in general, ML in domains with highly-structured data in particular. The track record of the team leaders, Pahikkala and Roos, in applied research already covers a large array of examples in which the preliminary steps of this research direction has been shown to be highly successful.

More information:

Teemu Roos, teemu.roos@cs.helsinki.fi

Tapio Pahikkala, tapio(dot)pahikkala(at)utu(dot)fi

Information, Complexity and Learning (ICL) research group

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Contact person:

Teemu Roos

Academy of Finland and NSF award funding for a joint project by HIIT and Carnegie Mellon University on ''Efficient and Robust Cognitive IoT Systems using Unreliable Sensors''

troos — Thu, 13 Apr 2017 13:44:38 +0000

The Academy of Finland and the National Science Foundation (NSF) have granted funding for a joint project by HIIT and the Carnegie Mellon University (CMU).

The project aims at providing a systematic understanding and designs of cross-layer cognitive IoT strategies that provide system-level resilience to faults in sensing and communication despite the unreliability and the limited field- of-view of the underlying IoT devices (as shown above). This understanding will be obtained by providing (i) fundamental performance limits; (ii) theoretical analyses of performance of designed strategies; and (iii) validation through realistic simulations of practical inference problems.

The project is titled "Efficient and Robust Cognitive IoT Systems using Unreliable Sensors", and the principal investigators are Assoc Prof Teemu Roos (HIIT) and Asst Prof Pulkit Grover (CMU). The two-year project belongs to the Wireless Innovation between Finland and US (WiFiUS) virtual institute.

More information:
Teemu Roos, teemu.roos@cs.helsinki.fi
Information, Complexity and Learning (ICL) research group

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Runner-up for ECAI 2016 Best Student Paper Award to Andreas Niskanen

mjarvisa — Tue, 06 Sep 2016 06:17:04 +0000

Andreas Niskanen was acknowledged as Runner-up for Best Student Paper Award at ECAI 2016, 22nd European Conference on Artificial Intelligence for the paper Synthesizing Argumentation Frameworks from Examples co-authored with Johannes P. Wallner and Matti Järvisalo. Currently Andreas is finishing his Master's thesis dealing with constraint optimization approaches to dynamic aspects of argumentation.

Together with IJCAI and AAAI, ECAI is a major biannual AI conference in the world.

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Paper "Pakota: A System for Enforcement in Abstract Argumentation" accepted to JELIA 2016

mjarvisa — Thu, 01 Sep 2016 05:42:26 +0000

The paper Pakota: A system for Enforcement in Abstract Argumentation by Andreas Niskanen, Johannes P. Wallner, and Matti Järvisalo, has been accepted for publication in the proceedings of the 15th European Conference on Logics for Artificial Intelligence (JELIA 2016).

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Paper "Causal Discovery from Subsampled Time Series Data by Constraint Optimization" accepted to PGM 2016

mjarvisa — Tue, 05 Jul 2016 07:50:56 +0000

The paper Causal Discovery from Subsampled Time Series Data by Constraint Optimization by Antti Hyttinen, Sergey M. Plis, Matti Järvisalo, Frederick Eberhardt, and David Danks has been accepted for publication in the proceedings of the 8th International Conference on Probabilistic Graphical Models (PGM 2016).

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Järvisalo to give invited talk at the Arg-LPNMR IJCAI-16 workshop

mjarvisa — Fri, 10 Jun 2016 15:16:24 +0000

Dr. Matti Järvisalo will deliver an invited talk at the First International Workshop on Argumentation in Logic Programming and Non-Monotonic Reasoning (Arg-LPNMR 2016) co-located with IJCAI-16 in New York City, USA, in July 2016.

Quoting the Arg-LPNMR website:

Research on argumentation and an Artificial Intelligence (AI) began in full force in the early eighties. The initial efforts showed how argumentation results in a very natural way of conceptualizing commonsense reasoning, appropriately reflecting its defeasible nature. In the mid-nineties, Dung (1995) has shown that argumentation provides a useful perspective for relating different non-monotonic formalisms. Currently, argumentation has been applied in different subfields of AI like Multi-Agent Systems, Semantic Web, Knowledge Representation and Reasoning, etc.

Argumentation has been revealed as a powerful conceptual tool for exploring the theoretical foundations of reasoning and interaction in autonomous systems and multiagent systems. Different dialogue models have been proposed based on the roots of argumentation. Indeed considering argumentation roots, the so called Agreement Technologies have been suggested in order to lead with the new requirement of interaction between autonomous systems and multiagent systems.

The aim of the workshop is to bring researchers from all these fields together, to present and discuss joint work, and identify possible shared future directions of the community.

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Järvisalo to deliver IJCAI-16 Early Career Spotlight as one of the most active early career AI researchers in the world

mjarvisa — Thu, 09 Jun 2016 12:36:05 +0000

Dr. Matti Järvisalo has been invited to deliver one of the Early Career Spotlight Talks at IJCAI-16, 22nd International Joint Conference on Artificial Intelligence taking place In New York City, USA, in July 2016. IJCAI is the prestigious number-one conference on artificial intelligence world-wide. The Early Career Spotlight Track at IJCAI-16 brings in---as some of the most active early career researchers in AI in the world---22 top talents in AI, most already holding professorships at world-leading research universities, as representatives of different modern branches of AI research to present on their individual research directions.

Quoting the official IJCAI-16 website:

Since its inception in 1969, IJCAI took great pride in representing all aspects of AI research. Even now, with a myriad specialized conferences, IJCAI remains uniquely suited to give the attendees an integrated snapshot of the whole of AI enterprise. In keeping with this goal, we are excited to announce a special track of "Early Career Spotlight" talks. The talks are aimed at providing an accessible introduction to the research directions of some of the most active early career researchers in all representative areas of AI.

The talks are by invitation, based on nominations from the IJCAI program committee, and will run in parallel to the technical sessions. It is our great pleasure to introduce the stellar speakers giving the Early Career Spotlight tracks at this leap year IJCAI.

The full list of IJCAI-16 Early Career Spotlight presenters is available here.

Add to highlights:

Importance:

Important

Show in:

HIIT news

CoSCo news

Paper "Impact of SAT-Based Preprocessing on Core-Guided MaxSAT Solving" accepted to CP 2016

mjarvisa — Thu, 09 Jun 2016 05:56:15 +0000

The paper Impact of SAT-Based Preprocessing on Core-Guided MaxSAT Solving by Jeremias Berg and Matti Järvisalo (Constraint Reasoning and Optimization group) has been accepted for publication in the proceedings of the 22nd International Conference on Principles and Practice of Constraint Programming (CP 2016).

Add to highlights:

Importance:

Normal

Show in:

CoSCo news

Two papers accepted to ECAI 2016

mjarvisa — Thu, 09 Jun 2016 05:42:22 +0000

The following two papers by the Constraint Reasoning and Optimization group have been accepted for publications in the proceedings of the 22nd European Conference on Artificial Intelligence (ECAI 2016) taking place in The Hague, The Netherlands, in August 2016.

Subsumed Label Elimination for Maximum Satisfiability by Jeremias Berg, Paul Saikko, and Matti Järvisalo.
Synthesizing Argumentation Frameworks from Examples by Andreas Niskanen, Johannes P. Wallner, and Matti Järvisalo.

Add to highlights:

Importance:

Normal

Show in:

CoSCo news