Strategic Partnership

Call 2014

Selected projects

Application of validated facial avatar paradigm to study of violence-exposed mothers with PTSD

Medicine

The use of facial avatars that one can transform or "morph" from one emotional expression to another, for example, speaking of high-arousal, negatively valenced affects such as from anxiety to anger and back, could help us test an important hypothesis concerning emotional differentiation before and after lab stressors. ln return, we as experts in child and adolescent psychiatry, we can help Todorov's lab with the design of more developmentally sensitive stimuli for different child-ages. Specifically, our hypothesis in the Geneva Early Childhood Stress Project is that maternai violence-related posttraumatic stress disorder (PTSD) impairs the reading of high-arousal, negatively valenced affects by mothers of their children's emotional expression (Schechter et al., 2006). We want to know if it also does so in their children in relation to their caregivers and peers.

This seed-funding grant initiative will permit a) the adaptation and mastery of the paradigm in relation to the Geneva Early Childhood Stress Project and b) will allow for further exchange that can foster an application for more substantive funding.

projects
Participants
Prof. Daniel Schechter
University of Geneva
Prof. Alexander Todorov
Princeton University

Democratic responsiveness and redistributive politics

Political Science

We seek funding to organize an exploratory workshop that would be held in Geneva in early June or September, 2014. The workshop would bring together political scientists from Princeton and UNIGE to explore the topic of democratic responsiveness in the context of rising socio-economic inequality. Why democratically-elected governments have not responded to rising inequality by engaging in compensatory redistribution is a "hot topic" in the study of American politics. Several Princeton faculty members (notably Martin Gilens, Nolan McCarty, and Thomas Romer) have contributed to this debate. Recently, students of European politics, including members of the department of political science at UNIGE, have begun to ask similar questions. The proposed workshop would serve as a venue for discussion of theoretical issues pertaining to the concept of "democratic representation" as well as new empirical research in this demain.
Our proposal builds on longstanding ties between the two departments. The workshop will provide an opportunity not only to explore common questions and potential collaborative research among the Princeton and UNIGE participants, but also to discuss how to organize a more regular, ongoing exchange of PhD students, and to raise funds for future workshops of a similar kind. We will also to invite researchers from the Graduate lnstitute (Cédric Dupont) and the University of Lausanne (Daniel Oesch and Jan Rosset).

projects
Participants
Prof. Jonas Pontusson
University of Geneva
Prof. Nolan McCarty
Princeton University

Enhanced superconducting wires for fusion applications

Physics

The proposed project is at the frontier between material science and engineering. The goal is to improve the understanding of the degradation mechanisms of superconducting cables and magnets to support the development of the next generation of thermonuclear fusion reactors. The two partners have advanced expertise in complementary demains, superconductivity and its applications for the group at UNIGE, and structural mechanics for the group at PPPL. This grant represents the main ingredient for a dynamic collaboration based on frequent exchanges of young scientists from the two laboratories.

This new connection between Princeton and UNIGE was just created this summer at the 23th International Conference on Magnet Technology when Dr. Zhai from PPPL met Prof. Senatore from UNIGE; both sides showed a strong interest in establishing a partnership collaboration to bring complementary technical expertise together to make impact to the field of designing better superconducting wires for fusion magnets.

projects
Participants
Prof. Carmine Senatore
University of Geneva
Dr. Yuhu Zhai
Princeton University

Small codes for large systems: information-theoretic privacy preserving content identification

Digital Studies
Informatics

The proposed project will focus on the information-­theoretic analysis of BOW based content identification systems. To our knowledge there is little work on the theoretical analysis of BOW‐systems' performance besides [5] and none on BOW based content identification. In addition, the privacy-preserving aspects of BOW based systems remain an emerging and little studied problem. In [6], it was shown that privacy-­‐preserving computations can be made efficient for specific functionals related to statistical estimation, and in [7] the construction of new polar coding schemes are shown to achieve the secrecy capacity with low complexity. The project will leverage these various techniques in the context of BOW based systems. The recent work on linear classification [8] will also be studied in this context.

The project applicants pose all necessary skills and have proven records of achievements to address the above problems. Therefore, the goal of this problem is to provide a simple and tractable model allowing to analyze, optimize and guide the design of BOW systems. We will consider the case of non-­‐compressed features to reveal the theoretical limits of BOW based identification systems, analyze the impact of descriptor compression and encoding/assignment as well as discovering the impact of geometrical consistency between the descriptors on overall system performance. Such a formulation was not considered in earlier studies.

References

  1. P. Moulin, “Statistical modeling and analysis of content identification,” in Information Theory and Applications Workshop (ITA), 2010. IEEE, 2010, pp. 1–5.
  2. A. L Varna and M. Wu, “Modeling and analysis of correlated binary fingerprints for content identification,” Information Forensics and Security, IEEE Trans-­‐ actions on, vol. 6, no. 3, pp. 1146–1159, 2011.
  3. F. Farhadzadeh, S. Voloshynovskiy, and O. Koval, “Performance analysis of content-­‐ based identification using constrained list-­‐based decoding,” IEEE Transactions on Information Forensics and Security, vol. 7, no. 5, pp. 1652–1667, 2012.
  4. S. Voloshynovskiy, O. Koval, F. Beekhof, F. Farhadzadeh, and T. Holotyak, “Information-­‐ theoretical analysis of private content identification,” in IEEE Information Theory Workshop, ITW2010, Dublin, Ireland, Aug.30-­‐Sep.3 2010.
  5. Yin Zhang, Rong Jin, and Zhi-­‐Hua Zhou, “Understand-­‐ ing bag-­‐of-­‐words model: a statistical framework,” Inter-­‐ national Journal of Machine Learning and Cybernetics, vol. 1, no. 1-­‐4, pp. 43–52, 2010.
  6. E. Abbe, A. Khandani, A. W. Lo, “Privacy-­‐preserving methods in systemic risk", Proc. and Papers of the American Economical Review (AER), 2012. New York Times article: http://bits.blogs.nytimes.com/2013/09/09/a-­‐data-­‐weapon-­‐to-­‐avoid-­‐the-­‐next-­‐financial-­‐crisis/
  7. E. Abbe, “Extracting randomness and dependencies via polarization, Slepian-­‐Wolf coding and secrecy” presented at ITA, Feb 2011, to appear in the IEEE Information Theory Transactions.
  8. E. Abbe, N. Alon, A. Bandeira, “Linear coding, classification and the space-­‐avoiding-­‐set problem”, preprint, 2013.
projects
Participants
Prof. Svyatoslav Voloshynovskyy
University of Geneva
Prof. Emmanuel Abbe
Princeton University

The Ising model in three and four dimensions

Mathematics

The proposed collaboration will connect a leading young researcher, of outstanding promise in his cohort, with an established authority
in the field. Thematically, the collaboration would link Modern Probability with related Mathematical Physics from which many of its deep challenges originate. It is expected to also form a basis for a network which has room for growth, seeding a collaboration on which a future European Research Council (ERC) grant application (by Hugo Duminil-Copin) may be based for a collaborative program which may grow to involve students and other young researchers, and possibly also other institutions. Princeton and Geneva Universities ossess exceptionally strong groups in the respective fields and the collaboration between the two principal investigators could extend to other members of these two research groups in order to tackle related problems.

Since its formulation by Lenz [Len20], the Ising model has been the most studied example of a system undergoing a phase transition.
The model was intended to explain the fact, discovered by Pierre Curie in 1895, that a ferromagnet loses its magnetization when heated above a critical temperature. Beyond its original motivation, the Ising model and its phase transition were found to be of a rather broad relevance, though of course its features do not exhaust the range of possible behaviors in statistical mechanics. The model also provided the testing ground for a large variety of techniques which have been later used to study other models. The goal of this proposal is to capitalize on the recent success (and the techniques developed) to pursuit the study of the Ising model.

projects
Participants
Prof. Hugo Duminil-Copin
University of Geneva
Prof. Michael Aizenman
Princeton University