Dong-Hee Kim's lecture on Mon 22 Aug: Dynamical mean-field theory
The dynamical mean-field theory (DMFT) is a non-perturbative method to solve correlated quantum systems in lattices, offering a robust tool to computationally study many-body phenomena in condensed matter physics. Here I present an overview of the methodology, covering the basic ideas, the formulations, and the technical aspects of computational implementations. Through self-consistent mapping of a full lattice problem onto a quantum impurity model, DMFT accurately describes local quantum fluctuations, allowing access to physics beyond usual static mean-field descriptions. Selected important numerically-exact methods to solve the quantum impurity model, including the exact diagonalization and quantum monte carlo methods, are sketched and compared to find their practical advantages and disadvantages. Some fundamental limitations, however, exist in the original DMFT as it neglects spatial quantum fluctuations. This can be lifted by the extension to the cluster DMFT that includes short-ranged nonlocal correlations. In addition, there is another variant, the real-space DMFT, which is still conventional but can consider the presence of a system-wide inhomogeneous potential in a single-particle level. I discuss briefly the applications of the real-space DMFT to the study of ultracold atomic Fermi gases that currently is one of hot topics in the condensed matter physics community.
Antti Puisto's lecture on Mon 22 Aug: Multiscale models for the rheology of thixotropic suspensions
Work by Antti Puisto, Xavier Illa, Mikael Mohtaschemi, Arttu Lehtinen, and Mikko J. Alava.
Abstract: A multiscale approach, based on a continuum level microrheological model for simulating the rheology of aggregating colloidal suspensions under rheometer conditions is presented. The model grounds on the simulation of colloidal aggregation using a well established continuum approach known as the Population Balance Equations (PBE). Through a semi-empirical constitutive equation the solved aggregate size distribution is connected to the suspension rheology. In the multiscale approach, a Couette rheometer is discretized radially to equally spaced shells, for which the local viscosity is calculated using the PBE method. The viscosity profile is used in the Navier-Stokes (NS) equations with appropriate boundary conditions to calculate the flow velocity and shear gradient profiles. Iterative solution of the PBE and NS allows tracing the time-evolution of the local rheology. The model is applied to understand experimental studies of transient shear banding (TSB) in aggregating colloidal suspensions demonstrating that the involved phenomenology is very similar to the case of simple yield stress fluids [1] and, as some of the present experiments show [2], observable only under appropriate boundary conditions.
[1] Divoux et al. Phys. Rev. Lett. 104, 208301 (2010).
[2] Besseling et al. Phys. Rev. Lett. 105, 268301 (2010).
Karoliina Honkala's lecture on Mon 22 Aug: Heterogeneous catalysis from first principles
Density functional theory calculations can be employed to various problems in physics and chemistry. In my presentation I will focus on some surface reactions relevant to heterogeneous catalysis. I'll discuss in detail how density functional theory calculations can be applied to study selective bond breaking and forming reactions for (chloro)ethenes over transition metal surface. Also the properties of ultra-thin MgO-film supported Au clusters and their reactivity in oxygen atmosphere will be presented.
Patrik Floreen's lecture on Wed 24 Aug: About EIT ICT Labs
The slides are here.
Professor Hiroshi Mamitsuka's lecture on Wed 24 Aug: Mining from Graphs
Abstract: Graphs are recently emerging and currently abundant data in a lot of domains, such as the internet, biology and chemistry. Graphs look highly complex, which makes a lot of important problems on mining from graphs still remain unsolved. In this talk, I will focus on two topics related with graphs: frequent pattern mining and clustering. First, a serious issue of frequent patterns from graphs is their redundancy. I will present a reasonable idea for summarizing the redundant patterns and an algorithm, which realizes this idea efficiently in the procedure of enumerating frequent patterns. I will further show the practical performance of this algorithm by applying the algorithm to real data of chemical compounds. Second, a practically well-seen setting on graphs is that given graphs share the same set of nodes and keep the common, unknown clusters, which appear only part of the given graphs. I will present a probabilistic model for this clustering problem and a variational Bayes learning algorithm for estimating parameters of this model from given graphs. I will confirm the performance of this algorithm by using synthetic graphs and gene networks, comparing with other possible approaches for this problem setting.
Student spotlight presentations on Wed 24 Aug
The students have been advised to have a 5-minute presentation (at most), consisting of 3 slides only: a title slide, a contents slide, and a slide saying what people in other fields should learn about it. It is important to consider the audience which consists of people in other fields of science. The order of the presentations is the following, going from physics to computational sciences and theoretical computer science to future internet:
Mikko-Heikki Mikkelä: Spectroscopy of free nanoclusters
Juha Koivisto: Complex systems and materials (TBC)
Vladimir Vagaytsev: Analytical-Numerical Methods for Localization of Hidden Oscillations in Dynamical Systems
Kirti Prakash: Combinatorial Patterns of Chromatin Modifications within a Nucleosome using Boolean Networks
Maria Osmala: Data fusion in epigenetics in predicting distal regulatory elements
Tuomo Sipola: Analyzing high-dimensional data
Ali Faisal: Pattern recognition and retrieval in huge repositories
Roland Kindermann: Symbolic Verification of Real-Time Systems
Mohammad Hoque (TBC)
Samu Varjonen: Secure Connectivity With Persistent Identities
Pekka Aakko's lecture on Thu 25 Aug: The start-up story of Applifier
Abstract: How to start a game company, fail few times and pivot the company to create the biggest cross promotion network for games on facebook - the story of Applifier.
Pauli Misikangas's lecture on Thu 25 Aug: Algorithms-as-a-Service – Business Opportunities for Algorithm Developers
Abstract: Algorithm development involves uncertainty and risks that are unbearable for most software companies. They cannot commit to R&D projects with unknown results, schedule, cost and applicability. Therefore, they rather go for the “keep-it-simple” solution. The fear of crossing this “Scientific Chasm” – the gap between a simple solution and an advanced algorithm ready for business use – leads to numerous lost business opportunities.
It is evident that there is an increasing need for smart data refining algorithms. Commercial success, however, requires cooperation between algorithm and application developers, a mutually profitable business model and a solid legal framework. The Cloud’N’Sci.fi algorithm marketplace based on the Algorithms-as-a-Service (AaaS) concept answers to these needs by enabling a smooth integration of algorithms into commercial applications, so that the non-scientific efforts required of the algorithm developer are minimized while his/her rights to IP and earnings remain protected. Commercializing algorithms does not prevent from continuing research or publishing the work – research and business shake hands!
Christian Queinnec's presentation on Thu 25 Aug: About EIT ICT Labs Doctoral School
The slides are here.
Antti Paasio's lectures on Thu 25 Aug: Entrepreneurship training
The slides are here
Mike P. Papazouglou's lecture on Fri 26 Aug: Research Roadmap for Service Oriented Computing
Abstract: Service-Oriented Computing is the computing paradigm that utilizes services as the basic constructs to support the development of rapid, low-cost and easy composition of distributed applications even in heterogeneous environments. The promise of Service-Oriented Computing is a world of cooperating services where application components are assembled with little effort into a network of services that can be loosely coupled to create flexible dynamic business processes and agile applications that may span organizations and computing platforms. The subject of Service Oriented Computing is vast and enormously complex, spanning many concepts and technologies that find their origins in diverse disciplines that are woven together in an intricate manner. The material in research spans an immense and diverse spectrum of literature, in origin and in character. As a result research activities are very fragmented. This necessitates that a broader vision and perspective be established—one that permeates and transforms the fundamental requirements of complex applications that require the use of the Service-Oriented Computing paradigm.
This talk provides a Service Oriented Computing Roadmap and places on-going research activities and projects in the broader context of this roadmap. In
particular, it fuses together elements of three computing paradigms based on the concept of services: Service Oriented Computing, Business Process
management and Cloud Computing. The research roadmap launches four pivotal, inherently related, research themes: service foundations, service
composition, service management and monitoring and service-oriented engineering.
About the speaker: Mike P. Papazouglou is Executive director of European Research Institute in Services Science (ERISS)
Alexandre Kandalintsev's lecture on Fri 26 Aug: Cloud-Aware Inter-Level Power Management
Abstract: In the last few years cloud computing has emerged as a popular means of addressing scalability problem of distributed large-scale applications. The ever-growing demand for such technology brought up the problem of efficient power management and resource allocation in the cluster. Several software and hardware technologies were developed to address these limitation. Software methods have shown to be quite efficient and can be applied to a whole cluster of hosts. However they do not have control over low-level hardware circuit modules. Built-in hardware methods have very fine-grained control, but their impact is limited to a specific microchip unit. In this study we seek to address this problem by developing algorithms that improve interoperability and combine the benefits of both software and hardware approaches delivering significant energy savings.
Keywords: Energy Efficiency, Clock Gating, Cluster Resource Management, Data Centers
Juwel Rana's lecture on Fri 26 Aug: Improving Group Communication by Harnessing Information from Social Networks and Communication Services
Abstract: The PhD thesis work investigates how the current trends in respect to mobility, social and context can be utilized in a coherent manner to improve group communication. More precisely, a social recommendation engine is proposed as a part of a larger framework that harnesses social media contents and on-line based interactions from social networks and communication services. To manage social contacts intelligently, the research aims at proposes weighted social graphs based on users communication patterns and context parameters. The main research questions are identified and addressed in some extent given below:
How to recommend a set of social contacts to form a group?
How can on-line interaction be modeled in a uniform manner to enable identification of users’ individual communication patterns?
How can personalized social graphs be built automatically by harnessing information from the social networking services and communication services?
How can a personalized service automatically calculate the social strength between contacts based on the interaction logs?
We consider this thesis as multidisciplinary where theories and technologies from different domains have been applied. For instance, graph theories for social relationship, Web data-mining techniques for social data analysis, clustering techniques for discovering groups, statistical analysis of communication history for calculating social strength, and different technologies of pervasive computing (e.g., Android OS based mobile apps with real-time context) and software engineering of distributed systems are applied in this work.
The current work investigates similarities and possible gaps between the on-line social networks and collaboration applications. It proposes a new platform of building collaborative social networks in a distributed fashion.
About the PhD Student: Juwel Rana is a PhD student at the Lulea University of Technology, Sweden. He works in Pervasive and Mobile Computing Division. He has MSc in Software Engineering of Distributed Systems form the Royal Institute of Technology (KTH), and Licentiate in Engineering Specialized in Media Technology from the Lulea University of Technology. His work is on group communication improvement considering current trend in Web-based communication, interaction, sharing and real-time collaboration.