CLASSIC is een jaarlijks evenement door studievereniging via, waarop verschillende masters uit de informatiewetenschappen aan de UvA zullen vertellen over de ervaringen, onderzoeken, en projecten die zij met hun master hebben gedaan. Hierdoor krijg je een indruk van wat er allemaal bij de verschillende masters komt kijken en wat zij met hun opgedane kennis allemaal hebben kunnen doen. Elke spreker zal ongeveer een half uur spreken, waarna er ruimte is voor vragen.
Inschrijving is niet nodig.
18:10-18:40 spreker 1: Jelte Fennema
18:45-19:15 spreker 2: Yfke Dulek
19:20-19:45 Pauze voor eten
19:45-20:15 spreker 3: Elise van der Pol
20:20-20:50 spreker 4: Jennifer Wennekers
20:55-21:25 spreker 5: Shabaz Sultan
Jelte Fennema (SNE)
Yfke Dulek (Master of Logic)
Elise van der Pol (Artificial Intelligence)
Jennifer Wennekers (BIS)
Shabaz Sultan (Computational Science)
Informatie over de sprekers:
Yfke Dulek zal een presentatie geven over haar project bij de Master of Logic.
Encrypting our data protects it from being read by others. The encrypted data (or "ciphertext") can then be sent over the internet, or stored on some device, without revealing information about the original contents, the "plaintext". In order to process the data, we usually have to decrypt the ciphertext back into the plaintext first. Homomorphic encryption is a form of encryption with a special property: it allows us to perform a computation on the (original) plaintext while only having access to the (encrypted) ciphertext. In order to evaluate a function on the plaintext, it suffices to evaluate some (related but different) function on the ciphertext. The result is an encryption of the output of the original function on the plaintext. This way we can securely outsource computations on our data: for example, some cloud server can perform all kinds of computationally demanding tasks on our data, and return the encrypted result to us without ever knowing the content of our data. Only recently, in 2009, the first fully homomorphic encryption algorithm was designed. In my thesis research, I consider homomorphic encryption for a different computational model: the quantum computer. Quantum computers seem to be more powerful in some ways than "classical" computers, and the data ("quantum bits" instead of just "bits") require new encryption techniques. What does it take to perform a quantum computation on encrypted quantum data? How can we devise an encryption that is strong enough so that attackers with quantum computers cannot break it? This research was conducted in part at the Centrum voor Wiskunde en Informatica (CWI) and QuSoft, the research center for quantum software. It is joint work with Christian Schaffner and Florian Speelman.
Yfke studied Cognitive Artificial Intelligence in Utrecht, and came to the UvA for the master of Logic. In her bachelor, she enjoyed foundational courses the most (such as modal logic, foundations of mathematics, and computational complexity). In the computation track of her master's she has continued along this line, learning about theoretical computer science, and is now working on quantum cryptography.
Elise van der Pol zal spreken over haar project bij de master Artificial Intelligence.
Reinforcement Learning is a variation of Machine Learning, where the task is not to assign labels to data points or cluster them together, but to find the optimal action to take in a given environmental state. To do so, we assign a Q-value to each (state, action)-pair, which is an expectation over the current and future reward from taking this action in this state. In most real world problems, the space of possible states is large and/or continuous, rendering it infeasible (or even impossible) to iterate over them. As such, we turn to Machine Learning models to approximate the value of each of the actions given our current state. With the recent popularity of Deep Learning, we have also seen great results of applying Deep Learning to Reinforcement Learning, as popularized by Deepmind in their recent success in playing Atari games and defeating Lee Sedol in Go (4-1!). In these applications, a Deep ConvNet is used to identify patterns in a complex state that are difficult to identify and extract manually. In my thesis, I look at the problem of traffic - networks of traffic lights- and minimizing the waiting time for cars. I apply the same concept of a Deep ConvNet to recognize patterns in traffic states to identify how agents - traffic lights - should switch their light states in order to maximize throughput. However, this problem is exacerbated by the fact that multiple locally optimal solutions do not necessarily lead to the global optimum. For example, to minimize local waiting time, a traffic light might want to allow a large number of cars to pass as fast as possible. However, this may lead to traffic jams further in the network, so that a globally optimal strategy could be to space the throughput of cars. The goal of my thesis is to employ coordination algorithms - such as max-plus - in order to find the perfect strategy for these cooperating traffic light agents.
My name is Elise van der Pol, a second-year student in UvA's Artificial Intelligence Master. I got my Bachelor's degree in Artificial Intelligence from UvA in 2014. I have volunteered for VIA a lot, having spent a year as chair of the board and many years as chair and member of various committees. During my Master's I have fallen in love with Reinforcement Learning, and I am currently in the process of doing my thesis research on the subject of coordination in deep reinforcement learning.
Shabaz Sultan zal spreken over zijn project bij de master Computational Science.
One of the fundamental problems in computational astrophysics is the so-called N-body problem, where multiple bodies interact with each other using gravity. There is no known general analytical solution for systems with more than two bodies. Instead we input initial conditions into a computational model and calculate the forces between bodies at a certain instant. Using integration methods that take these instantaneous forces as input the system can then be simulated over a period of time, so that we can know e.g. what the solar system or the Milky Way Galaxy will look like in the future.
In running these simulations we run into several problems. The integration methods simulate the system in discrete timesteps and will introduce inaccuracies. Also, these systems are sensitive to small perturbations in their initial conditions (i.e. they are chaotic) making it difficult to find a truly accurate solution of a system's behaviour over time. Using a special N-body code (devised by Tjarda Boekholt and Simon Portegies Zwart) we can however find true, converged solutions. This system uses an integrator that allows an integration to be run multiple times with increasingly small timesteps until a converging solution is found. This is combined with underlying calculations using arbitrary precision arithmetic, allowing the number of bits in the mantissa of numbers used during simulation to be increased as desired.
This system currently is quite slow, so the goal of my graduation project is to implement both the integrator and the arbitrary precision code on GPUs. In addition, to give insight into the integration methods and underlying code, an interactive simulation and visualisation system will be build that will allow students to explore both an N-body system and the workings of its underlying simulation code.
Shabaz Sultan is a second year Computational Science Master student, a joint programme from the UvA and the VU. He got his Bachelor's degree in Computer Science at the UvA, graduating in 2014 with a project building a system where using highly detailed data from laser altimetry the Netherlands can be interactively explored. His MSc graduation project is supervised by Prof.dr. Simon Portegies Zwart from the Leiden Observatory.
In moving from Computer Science to Computational Science -- an interdisciplinary programme -- he sought to get more of a connection with the natural sciences. Computational Science seeks to combine technical tools from computer science with domain knowledge from empirical sciences and numerical mathematical tools from applied mathematics to do research using computational models that would otherwise be hard or impossible using purely analytical or empirical tools.
Jennifer Wennekers zal spreken over haar project bij de Business Information Systems track van de master Information Sciences.
Abstract: During the master, I have participated in multiple (research) projects. Even though some projects were pre-defined concept wise, students were still allowed to give it their own twist, and personalise it to their own preferences and interests.
A summary of some of my performed projects:
Thesis: Compliance by Design - Governance for privacy of personal data for (C-)ITS.
Research regarding the governance of legislation around the privacy of personal data, that occurs in Cooperative/Connected Intelligent Transport Systems. I organised a working conference of two days with international experts from the legal and technical fields, representatives of EU commissions and the automotive industry. Next to that, I participated in writing a scientific paper, which will be published and presented at the ITS 2016 Congress in Glasgow in June. Project is still ongoing. Standardisation in IT Services: A Solution for Meeting Business Needs?
A Case Study of the standardisation programme of IT services at ABN AMRO.
Our team considered the concepts of ‘standardisation’, ‘agility’, ‘change’, ‘communication’ and ‘acceptance’ as the pillars of this case. These concepts shaped the base of the main research question: “How can communication enhance the agility of the IT standardisation services project?”
First, we wrote a literature review, to approach the concepts and to provide the academic background of the exploratory research. In the next phase, we conducted interviews to get insight on available perspectives of standardisation, agility and how communication is performed during the process within the organisation. Based on the results of the interviews, we presented our recommendations for the programme and indications of further research areas to ABN AMRO.
An Information Plan was created for EP Nuffic, containing proposed changes for EP Nuffic to improve the current application process of international students, who want to study in the Netherlands. The solution was agreed upon in co-operation with representatives of the UvA and IND. The goal of this project was to give the three parties new insights for their problem.
Elderly, Privacy, Behavioural Detection Systems and Smart Homes.
In this research, the impact of Behavioural Detection Systems located in smart homes, on the privacy of the elderly, was investigated. Smart homes for the elderly are seen as one of the ways to replace or to support for the care for the elderly in the future. With the perpetuating rise of health care costs, solutions in this field of work are widely investigated and considered a genuine possible solution to provide customised health care. However, many of the solutions do not consider privacy as a concern that should be taken care of by those who create the solutions, but rather by “others” (such as cryptographers). In smart homes, one of the major concerns is the lack of privacy by the ubiquity of sensors, which may yield in discovering the inhabitant’s behaviour in smart homes. To come to an understanding of how people look at these technologies in their homes, the research question that was proposed, was: “How can behaviour detection systems be implemented in smart homes for the elderly with a minimal impact on privacy?”.
Communication, Sense Making and Change Capacity in Organisations.
We performed a literature review in an attempt to find an answer to the main research question: “In what way can sensemaking be of a positive influence on the relationship between communication and change capacity within organisations?”. The reviewed concepts were ‘sensemaking’, ‘organisational change’, ‘communication’, and ‘change capacity’, and the existing relationships in between. Different scholar’s views were juxtaposed and contrasted in an attempt to connect the converging views they shared, or to mention the areas of expressed misalignment.
Sticks & Carrots: One or the Other?
Sticks and carrots, often used expressions for punishment and reward, have been part of the regulatory framework for years, if not centuries. Where sticks (punishment) were previously used to work as a lever to gain compliance from citizens, nowadays carrots (rewards) are seen as a more useful tool. Parenting can be seen as an example of this. “Modern” parenting involves rewarding the child for positive behaviour, in stead of punishing it for behaviour that is perceived as negative. Even though there is a trend of using carrots, some researches have proven that the stick is actually still a more useful tool to ensure compliance to regulations. Therefore, I compared sticks and carrots in this exploratory research, by discussing several researches. A preliminary conclusion was drawn regarding whether sticks or carrots should be preferred as a tool for compliance to governance.
Bio: My name is Jennifer Wennekers, and I am a student of the Business Information Systems track of the Information Sciences Master programme. BIS is the track that focusses on the alignment between business and IT, and examines how ICT can contribute to the realisation of organisational goals. Before my Master’s programme, I got my Bachelor’s degree in System and Network Engineering, with a specialisation in security and a focus on malware detection. Currently, I’m researching the governance of the privacy of personal data, which occur in Cooperative and Connected Intelligent Transport Systems, for my Master’s thesis.
Jelte Fennema zal spreken over zijn project bij de master System And Network Engineering.
A surprising number of services can be found on the Internet that are misconfigured in such a way that anyone can access them. This research shows how easy it is to misuse such services for malicious purposes, such as botnet command and control infrastructure. It also investigates the effect of the default configuration on the likelyhood that a service is misconfigured this way. This was done by gathering statistics about the number of real open occurences of different services on the Internet and comparing these against their default configurations. Finally recommendations will be given to developers on how to help prevent their software package from being affected by these issues.
Je moet ingelogd zijn om een formulier in te vullen.