1,721,111 research outputs found
Rete d'Eccellenza PASCAL
No full textL’obiettivo del progetto e’ stato quello di costruire una rete europea all’avanguardia nello studio di metodi di pattern analysis, modellazione statistica e learning computazionale fondati su impostazioni formali. Questi metodi sono impiegati come tecnologie di base per la progettazione di interfacce multimodali intelligenti, capaci di stabilire interazioni naturali con/fra esseri umani. In ogni fase del progetto l’uso di strumenti di Machine Learning e’ stato diimportanza cruciale: dal Machine Vision allo Speech, dallo Haptics alle Brain Computer In-
terface, dall’Information Extraction al Natural Language Processing, il Machine Learning ha offerto una metodologia unificante per l’integrazione multimodale
On multilabel classification and ranking with bandit feedback
No full textWe present a novel multilabel/ranking algorithm working in partial information settings. The algorithm is based on 2nd-order descent methods, and relies on upper-confidence bounds to trade-off exploration and exploitation. We analyze this algorithm in a partial adversarial setting, where covariates can be adversarial, but multilabel probabilities are ruled by (generalized) linear models. We show O(T1/2 log T) regret bounds, which improve in several ways on the existing results. We test the effectiveness of our upper-confidence scheme by contrasting against full-information baselines on diverse real-world multilabel data sets, often obtaining comparable performance
Improved lower bounds for learning from noisy examples: an information-theoretic approach
No full textAbstractThis paper presents a general information-theoretic approach for obtaining lower bounds on the number of examples required for Probably Approximately Correct (PAC) learning in the presence of noise. This approach deals directly with the fundamental information quantities, avoiding a Bayesian analysis. The technique is applied to several different models, illustrating its generality and power. The resulting bounds add logarithmic factors to (or improve the constants in) previously known lower bounds
Multiclass classification with bandit feedback using adaptive regularization
No full textIl file e' soggetto a copyright ed e' scaricabile da
http://www.springerlink.com/content/y6676227v7n17684
- …
