Pro-Multidis-CM


PROcesamiento MULTImedia DIStribuido
Programa S - 0505/TIC/0233, IV PRICIT, CM

Noticias Generales

warning: Creating default object from empty value in /mnt/datos/html/webs/doiras/public/modules/taxonomy/taxonomy.module on line 1388.

USER-CENTERED ISSUES IN RECOMMENDER SYSTEMS

Ponente: Dr. Nava TINTAREV (Telefónica I+D, Barcelona)

Recommender systems suggest items to purchase or examine based on users' preferences. In recent years there has have been a shift in what is considered important in recommender systems – from recommendation accuracy to other more user-centered criteria such as transparency, trust, satisfaction etc. In addition, traditional evaluation metrics such as root mean square error (RMSE) are increasingly being questioned.
In this talk, the speaker will give an overview of recommender systems, as well as related user-centered issues. Topics will include the speaker’s doctoral work on evaluations and explanations in recommender systems, and work currently being conducted at Telefónica R&D such as personalization for a travel scenario (within the scope of the WeKnowIt European project), as well as work with colleagues (X. Amatriain, J.M. Pujol, and N. Oliver) on how natural noise or inconsistency in user ratings affects errors in recommendation accuracy.

Fecha: 6 de Noviembre de 2009
Horario: 12:30 pm
Lugar: EPS Universidad Carlos III de Madrid, Campus de Leganés. Aula 4.3.A05, Edificio Torres Quevedo

Toward Cognitive Coexistence of Heterogeneous Wireless Users

Ponente: Prof. Lang TONG (SEE, Cornell University, Ithaca, NY, USA)

The explosive growth of wireless communications has presented us an intriguing spectrum paradox: On the one hand, the overcrowding spectrum allocation threatens the potential deployment of new wireless services. On the other hand, at a particular time and location, the spectrum is often only used sparsely, despite the presence of many potential unlicensed users. The root of this paradox, now widely recognized, is the static spectrum allocation and the rigid access policy that prevent a more efficient and dynamic usage of wireless channels.

We present in this talk some recent developments in opportunistic spectrum access using software-defined cognitive radios. We are interested in the coexistence of primary wireless users with cognitive secondary users capable of channel sensing and adaptive transmission. We focus on a decision-theoretic approach aimed at achieving an optimal tradeoff among opportunity learning, opportunity exploitation, and interference mitigation. We illustrate the proposed approaches in the context of the cognitive coexistence of Bluetooth/WLAN devices.

Fecha: 10 de Junio de 2009
Horario: 12:00
Lugar: EPS Universidad de Alcalá, Sala de Grados

Distributed Sensing and Inference in Random Information Fusion Networks

Ponente: Prof. Lang TONG (SEE, Cornell University, Ithaca, NY, USA)

Advances in microelectronics and wireless communication technology make it possible that a large number of sensors are networked to perform collaboratively tasks such as monitoring, learning, and computation. These sensors form a fusion network that allows sensors process their observations locally, share the information with other nodes, and extract information at fusion centers. The objective of an information fusion network is to extract useful information from these sensors in an efficient and economic fashion; there is a fundamental tradeoff between the cost of data fusion and the performance achieved at fusion centers.

This talk examines scalable fusion policies that achieve optimal inference at the fusion center and have a constant average cost per sensor as the size of the network increases. To this end, it is necessary to exploit statistical correlations among observations of sensor nodes. For statistical inference involving random dependency graphs, for example, we show that the sparsity of the dependency graph and that of the network graph play a crucial role of scalable information fusion. Some simple energy efficient fusion policies are presented.

Fecha: 09 de Junio de 2009
Horario: 10:00 - 11:00
Lugar: EPS Campus de Fuenlabrada de la Universidad Rey Juan Carlos, Edificio Departamental, Salón de Grados

Third Generation Machine Intelligence

Ponente: Prof. Christopher M. BISHOP (Ass. Director, Microsoft Research, Cambridge, UK)

The first successful applications of machine intelligence were based on expert systems constructed using rules elicited from human experts. Limitations in the applicability of this approach helped drive the second generation of machine intelligence methods, as typified by neural networks, which can be characterised as black box statistical models fitted to large data sets. In this talk I will describe a new paradigm for machine intelligence which has emerged over the last five years, and which allows strong prior knowledge from domain experts to be combined with machine learning techniques to enable a new generation of large-scale applications. The talk will be illustrated with example case studies.

Fecha: 25 de Mayo de 2009
Horario: 18:00
Lugar: EPS Universidad Carlos III, Edificio Padre Soler.

Probabilistic Machine Learning in Computational Advertising

Ponente: Dr. Joaquin Quiñonero

In the past few years online advertising has grown at least an order of magnitude faster than advertising on all other media. This talk focuses on advertising on search engines, where accurate predictions of the probability that a user clicks on an advertisement crucially benefit all three parties involved: the user, the advertiser, and the search engine. We present a fully probabilistic classification model that has the ability of learning from terabytes of web usage data. The model explicitly represents uncertainty allowing for fully probabilistic predictions.

Fecha: 25 de Mayo de 2009
Horario: 13:00
Lugar: EPS Universidad Carlos III, Edificio Torres Quevedo, Aula 4.3.A05

Distribuir contenido

Powered by Drupal - plantilla original: Artinet

Valid XHTML 1.0! Valid css!