Paris Dauphine University

About: Paris Dauphine University is a education organization based out in Paris, France. It is known for research contribution in the topics: Context (language use) & Population. The organization has 1766 authors who have published 6909 publications receiving 162747 citations. The organization is also known as: Paris Dauphine & Dauphine.

Bounded single-peaked width and proportional representation

Abstract: This paper is devoted to the proportional representation (PR) problem when the preferences are clustered single-peaked. PR is a "multi-winner" election problem, that we study in Chamberlin and Courant's scheme [6]. We define clustered single-peakedness as a form of single-peakedness with respect to clusters of candidates, i.e. subsets of candidates that are consecutive (in arbitrary order) in the preferences of all voters. We show that the PR problem becomes polynomial when the size of the largest cluster of candidates (width) is bounded. Furthermore, we establish the polynomiality of determining the single-peaked width of a preference profile (minimum width for a partition of candidates into clusters compatible with clustered single-peakedness) when the preferences are narcissistic (i.e., every candidate is the most preferred one for some voter).

Eliciting Information Concerning the Relative Importance of Criteria

Abstract: The notion of Relative Importance of Criteria (RIC) is central in the domain of Multiple Criteria Decision Aid (MCDA). It aims at differentiating the role of each criterion in the construction of comprehensive preferences, thus allowing to discriminate among pareto-optimal alternatives. In most aggregation procedures, this notion takes the form of importance parameters.

MDETR -- Modulated Detection for End-to-End Multi-Modal Understanding.

Abstract: Multi-modal reasoning systems rely on a pre-trained object detector to extract regions of interest from the image. However, this crucial module is typically used as a black box, trained independently of the downstream task and on a fixed vocabulary of objects and attributes. This makes it challenging for such systems to capture the long tail of visual concepts expressed in free form text. In this paper we propose MDETR, an end-to-end modulated detector that detects objects in an image conditioned on a raw text query, like a caption or a question. We use a transformer-based architecture to reason jointly over text and image by fusing the two modalities at an early stage of the model. We pre-train the network on 1.3M text-image pairs, mined from pre-existing multi-modal datasets having explicit alignment between phrases in text and objects in the image. We then fine-tune on several downstream tasks such as phrase grounding, referring expression comprehension and segmentation, achieving state-of-the-art results on popular benchmarks. We also investigate the utility of our model as an object detector on a given label set when fine-tuned in a few-shot setting. We show that our pre-training approach provides a way to handle the long tail of object categories which have very few labelled instances. Our approach can be easily extended for visual question answering, achieving competitive performance on GQA and CLEVR. The code and models are available at this https URL.

Combining UCT and Nested Monte Carlo Search for Single-Player General Game Playing

Abstract: Monte Carlo tree search (MCTS) has been recently very successful for game playing, particularly for games where the evaluation of a state is difficult to compute, such as Go or General Games. We compare nested Monte Carlo (NMC) search, upper confidence bounds for trees (UCT-T), UCT with transposition tables (UCT+T), and a simple combination of NMC and UCT+T (MAX) on single-player games of the past General Game Playing (GGP) competitions. We show that transposition tables improve UCT and that MAX is the best of these four algorithms. Using UCT+T, the program Ary won the 2009 GGP competition. MAX and NMC are slight improvements over this 2009 version.

Dark-energy constraints and correlations with systematics from CFHTLS weak lensing, SNLS supernovae Ia and WMAP5

Abstract: Aims We combine measurements of weak gravitational lensing from the CFHTLS-Wide survey, supernovae Ia from CFHT SNLS and CMB anisotropies from WMAP5 to obtain joint constraints on cosmological parameters, in particular, the dark-energy equation-of-state parameter w. We assess the influence of systematics in the data on the results and look for possible correlations with cosmological parameters. Methods We implemented an MCMC algorithm to sample the parameter space of a flat CDM model with a dark-energy component of constant w. Systematics in the data are parametrised and included in the analysis. We determine the influence of photometric calibration of SNIa data on cosmological results by calculating the response of the distance modulus to photometric zero-point variations. The weak lensing data set is tested for anomalous field-to-field variations and a systematic shape measurement bias for high-redshift galaxies. Results Ignoring photometric uncertainties for SNLS biases cosmological parameters by at most 20% of the statistical errors, using supernovae alone; the parameter uncertainties are underestimated by 10%. The weak-lensing field-to-field variance between 1 deg2-MegaCam pointings is 5-15% higher than predicted from N-body simulations. We find no bias in the lensing signal at high redshift, within the framework of a simple model, and marginalising over cosmological parameters. Assuming a systematic underestimation of the lensing signal, the normalisation increases by up to 8%. Combining all three probes we obtain -0.10 < 1 + w < 0.06 at 68% confidence ( -0.18 < 1 + w < 0.12 at 95%), including systematic errors. Our results are therefore consistent with the cosmological constant . Systematics in the data increase the error bars by up to 35%; the best-fit values change by less than 0.15.