CWI Scientific Meeting

This meeting will be a PhD meeting: six PhD students will present ongoing work from across CWI.
  • When Mar 31, 2017 from 01:00 PM to 02:00 PM (Europe/Amsterdam / UTC200)
  • Where Euler room Z009
  • Web Visit external website
  • Add event to calendar iCal

Dear colleagues,

I'm happy to announce the next CWI Scientific Meeting, to be held Friday, 31st of March, 13:00 in the Euler room (Z009). As usual, lunch will be provided beforehand.

This meeting will be a PhD meeting: six PhD students will present ongoing work from across CWI. See below for the program.

Hope to see you all there!


Tijs van der Storm
Peter van de Ven


## Srinivasan Arunachalam (AC)

Title: Strengths and weaknesses of quantum machine learning

There has been much recent interest in quantum algorithms for machine learning. These try to improve over classical algorithms for generalizing or otherwise learning from given data (the data itself could also be quantum). In this talk I will first introduce quantum machine learning, and then discuss a couple of positive and negative results in this area.

## Jasmijn Baaijens (LS)

Title: A puzzle with a million pieces: assembling viral genomes from
sequencing data

A virus infection usually consists of a group of closely related virus strains, together referred to as viral quasispecies. This diversity is the
result of extremely high mutation rates, which allows a viral population to rapidly adapt to its environment and evolve resistance to antiviral drugs. It is therefore of great importance to identify each of the individual genomes (haplotypes). However, this is a challenging task because of
sequencing errors and low strain abundance rates, especially in the absence of high-quality reference sequences. We present SAVAGE, a computational tool for reconstructing individual haplotypes of intra-host virus strains without the need for a reference genome. SAVAGE makes use of overlap graphs, where nodes represent sequencing reads, while edges reflect that two reads, based on statistical considerations, originate from the same virus strain. First, we use maximal cliques in the overlap graph to correct errors in the input sequences. Then, we apply an iterative scheme which extends the corrected sequences until individual haplotypes are found. In benchmark experiments on both simulated and on real datasets, SAVAGE outperforms state-of-the-art tools and we are able to reconstruct individual strains in Zika virus and hepatitis C virus patient samples.

## Krzysztof Bisewski (SC)

Title: How to minimize time-discretization error?

It is often of interest to find the tail of the distribution of a supremum of a real stochastic process over a finite time interval. When an explicit
expression for the distribution is unavailable it is usually approximated numerically. For most of the available methods the underlying process needs to be discretized in time. However, it requires a surprisingly dense discretization to achieve a negligible error. This in turn results in a
computationally expensive simulation. For the standard Brownian Motion we show that a naive discretization choice leads to the maximal possible error. We propose a solution to the problem using an adaptive placement of time-points.

## Tessel Bogaard (IA)

Title: Understanding user behavior from search logs: a metadata-level approach

Search log analysis is an unobtrusive technique for analysis of user behavior in digital systems. In a search context it contributes to an understanding of the information needs of users and to what extent these are met. Most studies focus on the query. This, however, suffers from various disadvantages. Firstly, queries form an uncontrolled vocabulary, can be ambiguous and provide little context. Secondly, most queries (and clicked documents) are long tail (i.e. they appear only once or twice in the logs), and as a consequence patterns of clicks and queries in a session have low support. Finally, queries contain highly privacy-sensitive information.

In digital libraries and archives, the collections are usually described with rich, curated metadata about their content. This metadata is often
reflected in so-called facets in the search interface to filter results. We propose a method of analysis that moves away from the query and focuses on this metadata, of both clicks and facets. We argue that a metadata-level analysis provides a good alternative to query-level analysis. It gives insights into long tail user behavior and information needs, and relieves some of the disadvantages mentioned above. Furthermore, it can be a starting point for inter-collection comparison of user behavior and is a first step towards more privacy-preserving method of analysis.

## Jan Willem Kleinrouweler (DIS)

Title: Delivering stable and high quality video over HTTP

Dynamic Adaptive Streaming over HTTP (DASH) is the dominant technology for online video streaming. Large content providers, such as YouTube and Netflix, together good for over 50% of internet traffic, implement DASH in their players. The technology has proven to be effective on the side of content providers. However, it exhibits major performance problems on the receiving end when DASH players share a network connection, or in network with heavy background traffic. The result is unstable and lower quality video. We present a so-called DASH Assisting Network Element. We implemented a Wi-Fi router that provides target bitrate signaling and dynamic traffic control. Trying to improve the streaming experience for users who share a network connection, our implementation increases the video bitrate and reduces the number of quality switches.

## Riemer van Rozen (SWAT)

Title: Live Game Design

In game development, continuously experimenting with game designs is essential for improving the gameplay quality. However, modifying digital
games is time-consuming due the representation gap between game designs and source code. A lack of powerful notations for modifying a game’s elements and timely feedback on adjustments prevents game designers from quickly experimenting with alternative game designs, and puts the game quality under pressure. We propose bridging the representation gap of game design with Live Game Design: a set of visual programming languages and rapid prototyping tools for modifying running games that provide immediate and continuous feedback and design suggestions to speed-up and focus the design process.