Helsinki Algorithms Seminar is a weekly meeting of researchers in the Helsinki area interested in the art of algorithms and algorithm design, broadly interpreted to cover both theoretical ideas and algorithm engineering on concrete computing platforms. In most cases we have a presentation prepared for each meeting to communicate an idea, a recent result, work-in-progress, or demo, but this should not be at the expense of discussion and simply having fun with algorithms.
Our affiliations are with Aalto University and the University of Helsinki, and accordingly our activities alternate between the Otaniemi Campus of Aalto University and the Kumpula Campus of University of Helsinki, catalyzed by the Helsinki Institute for Information Technology HIIT, under the Algorithmic Data Analysis (ADA) programme.
For the season programme, please see the seminar webpage.
**
Speaker: Mikaël Rabie
Title: Distributed Recoloring
Abstract:
Given two colorings of a graph, we consider the following question: can we re-color the graph from one coloring to the other through a series of elementary changes? Here, an elementary change, or step, consists in selecting an independent set and changing the color of each vertex to another also compatible with the colors of its neighbors.
The answer is not always positive, and it is in fact a PSPACE-complete decision problem.
In this talk, we try to quantify how considering a stable set at each step instead of a single vertex impacts the number of steps necessary for a re-coloring, and how much communication is needed in the LOCAL model (at each round, a vertex can only communicate with its neighbors, with no limitation as to the size of the messages nor the computation power of the vertex) to compute the re-coloring or decide there is none.
We also consider the question of how many extra colors are needed to make the problem always feasible, and beyond that, how the number of necessary steps decreases with the addition of colors. The case of trees is of special interest.
I will provide a collection of both positive and negative results around those questions.
This a joint work with Marthe Bonamy, Paul Ouvrard, Jukka Suomela and Jara Uitto.
]]>
Yves Vandermeer, Chair of the European Cybercrime Training and Education Group, will reveal the latest trends in cybercrime.
Cyber security is a trending topic nowadays, but what about cybercrime, dark markets and crypto currencies? Who are the cyber criminals, how are they organised, what tools do they use, and how do they choose their victims? Are we all targeted or are some of us more vulnerable? And finally, what is being done by law enforcement?
The lecture is open for all and free of charge, but requires registration in advance at https://haic-public-outreach.eventbrite.com.
Doors open at 17.30. The lecture will be approx. 45 min long after which there will time for questions and discussion.
This lecture is if the first in a series of public organised by Helsinki-Aalto Center for Information Security HAIC. The lectures are for anyone interested in the topics, no prior expert knowledge in computer science or related fields is required!
About the speaker: Yves Vandermeer holds an MSc in Computer Forensics, and has 20 years of experience in law enforcement as a computer crime and computer forensics practitioner. Since 2017, he has been working for the Norwegian Police University College where he carries out research on file systems forensics, live data and network forensics. His focus is on delivering knowledge and tools to law enforcement practitioners and improving computer crime fighting and computer forensics handling. As chairman of the European Cybercrime Training and Education Group, Yves promotes cooperation between Academic and LEA worlds, bringing topic experts together to raise expertise and address identified cyber training needs.
]]>Helsinki region machine learning researchers will start our week by an exciting machine learning talk. The aim is to gather people from different fields of science with interest in machine learning. Porridge and coffee is served at 9:00 and the talk will begin at 9:15.
Subscribe to the mailing list where seminar topics are announced beforehand.
Bayesian Deep Learning for Image Data
Abstract:
Deep learning is the paradigm that lies at the heart of state-of-the-art machine learning approaches. Despite their groundbreaking success on a wide range of applications, deep neural nets suffer from: i) being severely prone to overfitting, ii) requiring intensive handcrafting in topology design, iii) being agnostic to model uncertainty, iv) and demanding large volumes of labeled data. The Bayesian approach provides principled solutions to all of these problems. Bayesian deep learning converts the loss minimization problem of conventional neural nets into a posterior inference problem by assigning prior distributions on synaptic weights. This talk will provide a recap of recent advances in Bayesian neural net inference and detail my contributions to the solution of this problem. I will demonstrate how Bayesian neural nets can achieve groundbreaking performance in weakly-supervised learning, active learning, few-shot learning, and transfer learning setups when applied to medical image analysis and core computer vision tasks. I will conclude by a summary of my ongoing research in reinforcement active learning, video-based imitation learning, and reconciliation of Bayesian Program Learning with Generative Adversarial Nets.
Melih Kandemir
Professor of Computer Science, Özyeğin University
Dr. Kandemir studied computer science in Hacettepe University and Bilkent University between 2001 and 2008. Later on, he pursued his doctoral studies in Aalto University (former Helsinki University of Technology) on the development of machine learning models for mental state inference until 2013. He worked as a postdoctoral researcher in Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI) between 2013 and 2016. As of 2017, he is an assistant professor at Özyeğin University, Computer Science Department. Throughout his career, he took part in various research projects in funded collaboration with multinational corporations including Nokia, Robert Bosch GmbH, and Carl Zeiss AG. Bayesian deep learning, few-shot learning, active learning, reinforcement learning, and application of these approaches to computer vision are among his research interests.
**
See the next talks at the seminar webpage.
Please spread the news and join us for our weekly habit of beginning the week by an interesting machine learning talk!
Welcome!
]]>
14:00 – 14:30
Tuukka Tolvanen
Topic / Aihe: "Modeling protein-DNA binding specificities with random forest"
Supervisor / Valvoja: Prof. Emeritus Heikki Saikkonen
The focus of the seminar series is to highlight the research challenges and solutions faced by current and future information technology, as seen by the internationally leading experts in the field.
The vision of the series is to be approachable for audience that has scientific education not limiting to information technology, whilst at the same time providing information technology experts new viewpoints to their own discipline.
Venues alternate between University of Helsinki and Aalto University, the two host universities of HIIT.
Nic Lane, University of Oxford
The lecture is free of charge and open to everyone interested in the latest research in information technology. The lecture will be followed by an informal cocktail event. Please register here.
Abstract
In just a few short years, breakthroughs from the field of deep learning have transformed how computational models perform a wide-variety of tasks such as recognizing a face, driving a car or the translation of a language. Unfortunately, deep models and algorithms typically exert severe demands on local device resources and this conventionally limits their adoption within mobile and embedded platforms. Because sensor perception and reasoning are so fundamental to this class of computation, I believe the evolution of devices like phones, wearables and things will be crippled until we reach a point where current – and future – deep learning innovations can be simply and efficiently integrated into these systems.
In this talk, I will describe our progress towards developing general-purpose support for deep learning on resource-constrained mobile and embedded devices. Primarily, this requires a radical reduction in the resources (viz. energy, memory and computation) consumed by these models – especially at inference time. I will highlight various, largely complementary, approaches we have invented to achieve this goal including: binary “on-the-fly” networks, sparse layer representations, dynamic forms of compression, and scheduling partitioned model architectures. Collectively, these techniques rethink how deep learning algorithms can execute not only to better cope with mobile and embedded device conditions; but also to increase the utilization of commodity processors (e.g., DSPs, GPUs, CPUs) – as well as emerging purpose-built deep learning accelerators.
About the speaker
Nic Lane is an Associate Professor in the Computer Science Department at the University of Oxford. Before joining Oxford, he held dual appointments at University College London (UCL) and Nokia Bell Labs; at Nokia, as a Principal Scientist, Nic founded and led DeepX – an embedded focused deep learning unit at the Cambridge location. Nic specializes in the study of efficient machine learning under computational constraints, and over the last three years has pioneered a range of embedded and mobile forms of deep learning. This work formed the basis for his 2017 Google Faculty Award in machine learning. Nic’s work has received multiple best paper awards, including ACM/IEEE IPSN 2017 and two from ACM UbiComp (2012 and 2015). This year he will serve as the PC Chair of ACM SenSys 2018. Prior to moving to England, Nic spent four years at Microsoft Research based in Beijing as a Lead Researcher. He received his PhD from Dartmouth College in 2011. More information about Nic is available here.
**
For more information see the webpage of Helsinki Distinguished Lecture Series.
]]>
Christoph Lenzen
Max-Planck-Institut für Informatik
Host: Professor Jukka Suomela
Time: 14:15 (coffee at 14:00)
Venue: T3, CS building
Hazard-free Circuits
Abstract:
Computers and other hardware operate using the Boolean abstraction: signals are interpreted as logical 0 or 1, respectively, depending on whether their voltage is low or high. However, in some important cases, this abstraction is insufficient to accurately describe what happens in a circuit, as signals may take on intermediate values that are neither "clearly 0" nor "clearly 1." In such situations, a suitable abstraction is the ternary Kleene logic, which extends the operations of gates in a natural way by describing all "unclean" signals by a third symbol U. The goal is to control how U inputs affect the output of circuits, ideally ensuring (correct) Boolean outputs whenever possible. When a Boolean output could be achieved but a circuit does not guarantee this, this is referred to as a hazard of the circuit.
In this talk, I will give an overview of recent results (appearing at STOC'18) on avoiding hazards in circuits. In particular, I will discuss how monotone circuit lower bounds imply strong (unconditional!) lower bounds on the size of hazard-free circuits, for computational problems that permit small circuits (with hazards). On the positive side, I show general transformations that derive hazard-free circuits from arbitrary circuits. The talk will be accessible to a general CS audience; in particular, prior knowledge on circuit complexity is neither assumed nor needed (which is fortunate, as the speaker does not possess it either).
Short bio:
Christoph Lenzen received a diploma in mathematics from the University of Bonn and a Ph.D. from ETH Zurich. After postdoc positions at the Hebrew University of Jerusalem, the Weizmann Institute of Science, and MIT, he took on his current position as group leader at MPI for Informatics, Saarbrücken, Germany. His research interests span from the theory of distributed systems to designing fault-tolerant hardware. An ERC starting grant supports his work in the latter area, which lead to the results presented in this talk.
]]>Onerva Korhonen, M.Sc. (Tech) will defend the dissertation: “The quest for consistency: Effects of node definition and preprocessing on the structure of functional brain networks”.on 2 March 2018 at 12 o'clock at the Aalto University School of Science. In the dissertation, the candidate investigated methodological questions related to mapping the functional networks of the human brain. The results of the dissertation show that methods commonly used in network neuroscience may affect the results of network analysis in an unexpected way.
Network neuroscience models the human brain as a complex network, or as a system of nodes and links representing connections between the nodes. Although the network model has greatly increased our understanding on the structure and function of the human brain, a number of methodological questions of network neuroscience still remain unanswered. This dissertation concentrates on two of the questions: how should one define nodes of functional brain networks and how does preprocessing of the data affect the structure of the obtained networks. The results of the dissertation show that commonly used methods of network neuroscience have unexpected effects on the structure of the obtained networks. Further, these methods are partially based on insufficiently justified assumptions.
Regions of Interest (ROIs) are commonly used as nodes of functional brain networks. These ROIs consist of several measurement voxels of functional magnetic resonance imaging (fMRI) or source space vertices of electroencephalography (EEG) and magnetoencephalography (MEG). The ROI approach is based on the assumed functional homogeneity: each voxel or vertex of an ROI should behave similarly in time. However, according to the results of the dissertation, this assumption is not true for several sets of ROIs commonly used in fMRI, EEG, and MEG studies. Therefore, it is questionable if these ROIs are reasonable nodes for functional brain networks. Further, both functional homogeneity of ROIs and the local structure of functional brain networks change in time, which highlights the dynamic nature of functional brain networks. One may well ask if the optimal network model of the human brain can be based on any static set of nodes.
Making use of the full potential that methods of network science offer for neuroscience requires a solid methodological basis. The present dissertation wishes to warn about the consequences of careless methodological choices and to remind about how important the continuous, careful methodological work is for network neuroscience.
Dissertation release (pdf)
Opponent: Associate Professor Javier Martín Buldú, Center for Biomedical Technology, Madrid, Spain
Custos: ProfessorJari Saramäki, Aalto University School of Science, Department of Computer Science
]]>
Joachim Spoerhase
Universität Würzburg, Germany
Host: Prof Parinya Chalermsook
Time: 14:15 (coffee at 14:00)
Venue: T5, CS building
Constant-Factor Approximation for Ordered k-Median
Abstract:
We study the Ordered k-Median problem, in which the solution is evaluated by first sorting the client connection costs and then multiplying them with a predefined non-increasing weight vector (higher connection costs are taken with larger weights). The problem unifies many fundamental clustering and location problems such as k-Median and k-Center. This generality, however, renders the problem intriguing from the algorithmic perspective. Recently, Aouad and Segev proposed a sophisticated local-search based O(log n) approximation algorithm for general weight vectors, extending the result by Tamir (2001) for the case of a rectangular weight vector, also known as k-Facility p-Centrum. The existence of a constant-factor approximation algorithm remained open, even for the special case with a rectangular weight vector.
Our main result is an LP-rounding constant-factor approximation algorithm for the Ordered k-Median problem with general weight vectors.
We first provide a new analysis of the rounding process by Charikar and Li (2012) for k-Median, when applied to a fractional solution obtained from solving an LP with a carefully modified objective function, results in an elegant 15-approximation for the rectangular case. In our analysis, the connection cost of a single client is partly charged to a deterministic budget related to a combinatorial bound based on guessing, and partly to a budget whose expected value is bounded with respect to the fractional LP-solution. This approach allows to limit the problematic effect of the variance of individual client connection costs on the value of the ordered objective function of the Ordered k-Median problem. Next we analyze objective-oblivious clustering that allows to handle multiple rectangles in the weight vector and obtain constant factor approximation for the case of O(1) rectangles. Then, we show that a simple weight bucketing can be applied to general weight vectors resulting in O(log n) rectangles and hence in a constant factor approximation in quasi-polynomial time. Finally, with a more involved argument, we show that also the clever distance bucketing by Aouad and Segev can be combined with the objective-oblivious version of our LP-rounding for the rectangular case, and that it results in a true, polynomial time, constant-factor approximation algorithm.
This work was accepted to STOC 2018.
]]>Kari Kostiainen
ETH Zurich
Host: Professor N. Asokan
Time: 14:15 (coffee at 14:00)
Venue: T3, CS building
PRCash: Centrally-Issued Digital Currency with Privacy and Regulation
Abstract:
Decentralized cryptocurrencies based on blockchains provide attractive features, including user privacy and system transparency, but lack active control of money supply and capabilities for regulatory oversight, both existing features of modern monetary systems. These limitations are critical, especially if the cryptocurrency is to replace, or complement, existing fiat currencies. Centralized cryptocurrencies, on the other hand, provide controlled supply of money, but lack transparency and transferability. Finally, they provide only limited privacy guarantees, as they do not offer recipient anonymity or payment value secrecy.
In this talk we introduce a novel digital currency, called PRCash, where the control of money supply is centralized, money is represented as transactions for transferability and improved privacy, and transactions are verified in a distributed manner and published to a public ledger for verifiability and transparency. Strong privacy and regulation are seemingly conflicting features, but we overcome this technical problem with a new regulation mechanism based on zero-knowledge proofs. Our implementation and evaluation shows that payments are fast and large-scale deployments practical. PRCash is the first digital currency to provide control of money supply, transparency, regulation, and privacy at the same time, and thus make its adoption as a fiat currency feasible.
]]>Welcome to the annual Get-together for doctoral students at the School of Science!
This year the theme is: The Ups And Downs of Doctoral Studies.
Come and hear different stories about what doctoral studies are like!
When: 9 March 2018, at 9:00-11:00 (breakfast brunch is served from 8.30 onwards)
Where: The Undergraduate Centre, Otakaari 1X, room A235, 2^{nd} floor (see the map of Otaniemi here and for the building and entrance here)
Programme:
For breakfast orders, please sign up by 1 March: https://www.webropolsurveys.com/S/4CCCCF686E7138B5.par
Please note that registration is binding.
]]>The popular Aalto Fintech Seminar Series will continue in March - April 2018 providing a holistic view of the financial industry and its new challenges and opportunities stemming from digitalisation and new technologies.
Finnish top experts will share latest knowledge on topical subjects such as blockchains, machine learning and artificial intelligence, big data and data analytics, data privacy and cyber security. The rapidly changing global landscape with new Fintech start-ups and large tech companies, and new regulation will also be discussed.
The focus will be on practical questions and solutions. You may join all lectures or pick one or two. The seminar is free-of-charge and open to everyone. Welcome!
Enrolment: https://www.webropolsurveys.com/S/745136449DDC82E5.par
Students: 1-3 credits available for students (enrolment in WebOodi). MyCourses
The seminar will take place in TUAS building, Lecture Hall TU2, Maarintie 8, Otaniemi Campus | Map
Wednesday 14.3. at 16.15-17.45
Peter Sarlin, Hanken, Silo.AI: Presentation on Artificial Intelligence in Finance
Päivi Heikkinen, Bank of Finland: Do you read Crypto?
Wednesday 21.3. at 16.15-17.45
Markus Hautala, Tieto: Sovrin Global Identity Network – Solving the Trust Issues of the Digital Age
Hanna Lankinen, OP Finance Group: Presentation on Data Privacy
Wednesday 28.3. at 16.15-17.45
Timo Ritakallio, Ilmarinen: Banking Industry and Digital Disruption
Kevin Linser, Selma: What worked & what didn’t – Behind the Scenes of a Fintech Startup
Wednesday 11.4. at 16.15-18.00
Antti Kiuru, Nordea: Presentation on Cybersecurity
Kim Ristolainen, Bank of Finland: Presentation on Artificial Intelligence and Financial Crises
Antti Kosunen, Nestholma: Learnings from Fintech / Bank Collaboration
Wednesday 18.4. at 16.15-17.45
For more information: fintech.aalto.fi or ruth.kaila@aalto.fi, Department of Industrial Engineering and Management, Aalto University
]]>Jussi Korpela, M.Sc. (Tech.) will defend the dissertation: "Exploring correlated data: confidence bands and projections of shared variation" on 15 March 2018 at 12 noon at the Aalto University School of Science.
Opponent: Professor Mykola Pechenizkiy, Eindhoven University of Technology, The Netherlands
Custos: Professor Kai Puolamäki, Aalto University School of Science, Department of Computer Science
]]>
Secure Systems Group organizes every summer a “Demo Day” to showcase their work during the previous 12-month period (link to the previous Demo Days).
Next Demo Day will be held on June 20, 2018 in CS building. This event is jointly organized by the Secure Systems Group at Aalto University and the University of Helsinki. The Demo Day provides an opportunity for Secure Systems Group to present their work and seek valuable inputs from external visitors. This is a public event and everyone is welcome to attend.
Please register through this link >> registration form
See the event webpage for more information.
Preliminary program | |
---|---|
12:00-12:10 | “Welcoming words” by Prof. N. Asokan |
12:10-12:55 | “State-of-the-Union” presentation by Prof. N. Asokan |
13:00-17:00 | Demos and poster presentations |