Video about Sune

The amazing Villum Foundation (who first funded the SensibleDTU project back in 2012) wanted to showcase some of the work coming out of their Young Investigator Program. As part of that, they have produced a beautiful video about my research (and also a bit about who I am).

It’s directed by Ole Stenum who shot something like 5 hours of conversation and elegantly captured the essence the whole thing in a 3 minute film. It was fun to try to be part of the whole process.


Benjamin Maier Talk

This Wednesday we will have Benjamin Maier, a PhD student (physics) from Dirk Brockmann’s group speaking at DTU. Ben is based at the Robert Koch-Institut but also affiliated with Humboldt University Berlin’s physics department and IRI Life Sciences. His interest lies in identifying the underlying processes of human contact facilitating the spread of diseases. In particular he is investigating the influence of structural properties of human contact networks or human transport networks, both of static and dynamic nature.

  • Title: Flockworks, A class of dynamic network models for face-to-face interactions
  • Date: Wednesday, November 29th
  • Time: 13:30
  • Location:  DTU, Building 321, 1st floor lab space

Abstract: Studying the dynamics of face-to-face interaction networks is essential for a better understanding of contact mediated processes, contagion processes, and disease spreading. In many studies regarding social systems, networks are reconstructed using time averages or integrated networks, in which links reflect an interaction likelihood, although frequently this measure is not well defined but serves as a qualitative feature from which network properties are computed. During the last years a significant effort was made to resolve this issue by developing algorithms to analyse dynamic processes on the actual time-dependent contact patterns of social systems. However, there is still a lack of simple dynamic network models generating temporal networks of typical behaviour observed in real systems.

We introduce a class of minimal dynamic network models that naturally yield group formation and are easy to control. In those models, randomly chosen individual nodes cut their existing links, connect to a target node and establish links to the target’s neighbors. We discuss a variety of properties of those models and show how to use it for comparison of epidemic processes on real-world data.

Sune Lehmann Talk

Next week, I’ll be giving a talk about SensibleDTU at DTU. I hope the text below explains everything.

  • Title: Sensible DTU. Is that project still going on? If yes, I wonder what the h&ck they’re working on these days?
  • Date: Tuesday Nov 7th, 2017.
  • Time: 13:00
  • Location: DTU, Building 324, Room 030

Abstract: We’ve actually been doing a lot of interesting work on the Sensible DTU dataset  over the past year or so. (SensibleDTU is the project where we collected + dynamic multilayer network and behavioral data from 1000 smartphones) . This talk covers highlights and goes in depth with the most exciting projects. And you have a chance to ask questions: Maybe there’s something in there for you to test your own algorithm on. I’m also considering delivering some deep and personal revelations.

Notice: The talk will be filmed as part of a movie created by the Villum Foundation, who funded a large chunk of Sensible DTU. And – if you agree – you might end up in the film.

Hartmut Lentz Talk

For the concluding speaker in our September-talks series, I am happy to announce that Hartmut Lentz will talk about his work on spread of infectious disease on temporal networks. Hartmut works at the “Institut für Epidemiologie” at the Friedrich Loeffler Institute. He is a fantastic speaker and an authority on temporal networks. Full details below.

  • Date: Thursday, September 28th
  • Time: 1pm
  • Location: DTU, Building 321, 1st floor lab-space
  • Title: Spread of infectious diseases in temporal networks

Abstract: Many networks are treated as static objects, although they are in fact strongly time-dependent. This can have a dramatic impact on the possible spreading patterns for infectious diseases.

A static (aggregated) trade network is constructed as follows: if two nodes are connected directly to each other in a time-dependent network, the same connection is present in the static network. A fundamental difference between the static and the time dependent view however, is the consideration of paths, i.e. indirect connections over more than one edge. Concerning paths, the causality of the edges used plays an essential role. In an aggregated network, paths can seem causal, although they do not follow a time-respecting sequence of edges in the real system. This leads to a systematic overestimation of outbreak sizes, if time-dependent networks are treated as static.

We introduce a new method, which allows for the computation of the total causal path structure of a temporal network (represented by its accessibility graph) using the adjacency matrices of its snapshots. In addition, information about the timescales required for path traversal can be derived from the step-by-step derivation of the accessibility graph of the network. This procedure directly yields the distribution of shortest path durations in a temporal network. In addition, we define the new measure causal fidelity that compares the number of paths in a temporal network with its aggregated counterpart. This measure allows a quantitative assessment of how well a temporal network can be approximated by a static aggregated one.

The methods presented here require only basic knowledge linear algebra and can be implemented efficiently. Their capability is demonstrated for three examples: networks of social contacts, livestock trade, and sexual contacts.

Philipp Lorenz Talk

The September talk series continues full steam ahead. This week, you have a chance to see Philipp Lorenz talk about dynamics of topics in online social media. Philipp is a PhD student at TU Berlin’s Institut für Theoretische Physik in the Nonlinear Dynamics and Control: empirical networks and neurodynamics group. Phillip’s work focuses on temporal communities of hashtags, modeling the rise and fall of online topics, threshold models with repost and recovery, and more. Details of the talk below

  • Date: Tuesday, September 19th
  • Time: 2pm
  • Location: DTU, Building 321, 1st floor lab-space
  • Title: Capturing and modeling the dynamics of online topics

Abstract: Online media have a huge impact on public opinion, economics and politics. Every day, billions of posts are created and comments are written, covering a broad range of topics. Especially the format of hashtags, as a discrete and condensed version of online content, is in our focus. Here we present a pipeline, consisting of methods from static community detection as well as novel approaches for tracing the dynamics of topics in temporal data. We build co-occurrence networks from hashtags with timestamped edges. On static snapshots we infer the community structure and solve the resulting bipartite matching problem, by taking into account higher order memory. The results are robust to temporal fluctuations and instabilities of the static community detection.

The resulting dynamics in various datasets and for different observables, such as the community sizes or the likes they gather, as a proxy for the popularity of a topic, we observe universal behavior. Despite their versatility we find that in all datasets the distributions of gains and losses in popularity are fat-tailed, indicating occasional but large and sudden changes in public interest.

We hypothesise that only a few mechanisms may govern this behavior:

  • Gaining interest follows the rule of preferential attachment .
  • Saturation of the limited attention span decreases its fame.
  • discrete ranking leads to a competition between threads.

With these ingredients, we are able to design a class of models, which can reproduce the qualitative dynamics and the quantitative distributions of dynamical properties in the empirical observations. The model parameters and the required configuration for a given dataset is informational with respect to the sociological and psychological mechanisms that drive the dynamics of popularity in different contexts.

Bjarke Felbo on Emoji and Emotions

We continue the streak of exciting September talks. This time it’s DTU alum (now MIT) Bjarke Felbo who recently caused international press frenzy (see MIT Technology Review, BBC, Newsweek, Business Insider, The Telegraph, The Register, Huffington Post (FR), Numerama (FR) for details) with his sarcasm-savvy deep learning algorithm. Now there’s a great opportunity you can get all the technical details and ask questions, etc.

  • Place: Technical University of Denmark, Building 210, room 112.
  • Date: September 13th, 2017
  • Time: 13:00

Title: Using millions of emoji occurrences to learn any-domain representations for detecting sentiment, emotion and sarcasm
Abstract: NLP tasks are often limited by scarcity of manually annotated data. In social media sentiment analysis and related tasks, researchers have therefore used binarized emoticons and specific hashtags as forms of distant supervision. Our paper shows that by extending the distant supervision to a more diverse set of noisy labels, the models can learn richer representations. Through emoji prediction on a dataset of 1246 million tweets containing one of 64 common emojis we obtain state-of-the-art performance on 8 benchmark datasets within sentiment, emotion and sarcasm detection using a single pretrained model. Our analyses confirm that the diversity of our emotional labels yield a performance improvement over previous distant supervision approaches.

Piotr Sapieżyński on Gender and Fairness

Friend of the lab & postdoc at Northeastern University Piotr Sapieżyński is visiting Copenhagen and we’re lucky to hear about his ongoing work on FAT (Fair, Accountable, and Transparent) Machine Learning. This talk which focuses on the fair part of FAT ML is not one to miss if you want to be on the cutting edge of ethically responsible Machine Learning.
  • Date: September 7th, 2017
  • Time: 13:00
  • Place: Technical University of Denmark, Building 321, first floor lab space.
Title: Academic performance prediction in a gender-imbalanced environment
Abstract: Individual characteristics and informal social processes are among the factors that contribute to a student’s performance in an academic context. Universities can leverage this knowledge to limit drop-out rates and increase performance through interventions targeting at-risk students. Data-driven recommendation systems have been proposed to identify such students for early interventions. However, we find that the performance of some students is best predicted using indicators that differ from those predictive for the majority. Naive approaches that do not account for this fact might favor the majority class and lead to disparate mistreatment in the case of minorities. In this presentation I will talk about behavioral and psychological differences between male and female participants of the Copenhagen Networks Study, and how these differences can contribute to unequal performance in the academic achievement prediction problem. I will also stress the importance of the error analysis in seemingly well-performing predictors and review the approaches to fair machine learning.

Bernardo Huberman Visit

Wow. We are lucky to have legendary researcher Bernardo Huberman visiting later this month. His production of high-impact papers, books, and patents are is too rich and plentiful to reproduce here, so I’ll simply quote Wikipedia’s summary!

Bernardo Huberman is a Senior Fellow and Senior Vice President at Hewlett Packard Enterprise Company, and Director of the Mechanisms and Design Lab at Hewlett Packard Labs. He is currently a Consulting Professor in the Department of Applied Physics and the Symbolic System Program at Stanford University.

Bernardo has been a central player throughout the rise of network theory (and mentor for field notables, such as Lada Adamic and Jure Leskovec), but that’s just a fraction of what he’s accomplished. If you care about anything related to information sciences, this is a talk you cannot miss. Here are the details:

  • Date: August 29, 2017.
  • Time: 14:00
  • Location: Technical University of Denmark, Building 324, Room 040.

Title: Social media and the attention economy

Abstract: We are witnessing a momentous transformation in the way people interact and exchange information with each other. Content is now co-produced, shared, classified and rated by millions of people, while attention has become the ephemeral and valuable resource that everyone seeks to acquire. This content explosion is to a large extent driven by a mix of novel technologies and the deep human drive for recognition. This talk will describe the regularities that govern how social attention is allocated among all media and the role it plays in the production and consumption of content. It will also describe how its dynamics determines the emergence of public agendas while allowing predict the evolution of social trends.

Roberta Sinatra visit

I’m very excited to have Roberta Sinatra visiting the group for the week of April 3rd. She is an Assistant Professor at the Center for Network Science and Math Department at the Central European University in Budapest.

Roberta works on ‘the science of success’, her most recent adventures resulting in two very impressive pieces in the interdisciplinary journal Science (and corresponding world wide press coverage). Check out those papers here and here.

She will give a talk about her work at DTU Compute. Details can be found below.

  • Date: April 4th, 2017
  • Time: 13:00
  • Location: Technical University of Denmark, Building 321, 1st floor lab space.

Title: Quantifying the evolution of individual scientific impact

Abstract:Despite the frequent use of numerous quantitative indicators to gauge the professional impact of a scientist, little is known about how scientific impact emerges and evolves in time. In this talk we quantify the changes in impact and productivity throughout a career in science and show that impact, as measured by influential publications, is distributed randomly within a scientist’s sequence of publications. This random impact rule allows us to formulate a stochastic model that uncouples the effects of productivity, individual ability and luck, unveiling the existence of universal patterns governing the emergence of scientific success. The model assigns a unique individual parameter Q to each scientist, which is stable during a career and accurately predicts the evolution of a scientist’s impact, from the h-index to cumulative citations. Finally, we show that the Q-parameter is more predictive of independent recognitions, like prizes, than cumulative citations, h-index or productivity.

Michael Szell Visit

We are very lucky to have Michael Szell visiting the week of April 3rd. Micheal is a research fellow at the Hungarian Academy of Sciences, Centre for Social Sciences and visiting at Northeastern University, Center for Complex Network Research. He’s previously worked at the MIT Media Lab’s Senseable City Lab.

Michael’s research focuses on a quantitative understanding of collective behavior. How the the underlying patterns of our interlinked actions and decisions can be modeled in computational social science, and his past research involves mining and modeling large-scale data sets of human activity following a complex networks/systems approach.

His exciting work has been featured in PNAS, Nature Physics, Science, and many other fine journals. During his visit, Michael will give a talk at DTU Compute.

  • Date. Tuesday April 4th, 2017
  • Time: 14:00
  • Location. Technical University of Denmark, Building 321, 1st floor lab space.
Title: Using network science and data visualization to assess the potential of urban sharing  economies
Abstract: We introduce the notion of shareability network, which allows us to model the collective benefits of sharing rides as a function of passenger inconvenience, and to efficiently compute optimal sharing strategies on massive datasets. We first apply this framework to a dataset of millions of taxi trips taken in New York City, showing that with increasing but still relatively low passenger discomfort, cumulative trip length can be cut by 40% or more. This benefit comes with reductions in service cost, emissions, and with split fares, hinting toward a wide passenger acceptance of such a shared service. Simulation of a realistic online system demonstrates the feasibility of a shareable taxi service in New York City. Shareability as a function of trip density saturates fast, suggesting effectiveness of the taxi sharing system also in cities with much sparser taxi fleets or when willingness to share is low. Indeed, applying the same framework to a diverse set of world cities, using data on millions of taxi trips beyond New York City, in San Francisco, Singapore, and Vienna, we compute the shareability curves for each city, and find that a natural rescaling collapses them onto a single, universal curve. We explain this scaling law theoretically with a simple model that predicts the potential for ride sharing in any city, using a few basic urban quantities and no adjustable parameters. Accurate extrapolations of this type will help planners, transportation companies, and society at large to shape a sustainable path for urban growth. Finally, we present “What the Street!?”, an online platform for the interactive exploration of city-wide mobility spaces, published in April 2017. The aim of What the Street!? is to facilitate the intuitive exploration of (wasted) mobility space in cities, exploring why and to which extent space is distributed unevenly between different modes of transportation. We demonstrate how this data visualization of re-ordered city spaces can effectively inform relevant stakeholders and the public about large-scale reductions of parking spaces in future scenarios of wide-spread car-sharing.