Grant: Strategic Projects Supplemental Competition
Université de Montréal
$99,000.00
Deep learning of similarity spaces for sequential web data
EF 2007-08
Résumé
Recent theoretical and empirical work in statistical machine learning has demonstrated the importance ofdeveloping learning algorithms for deep architectures, i.e., function classes obtained by composing multiplenon-linear transformations. These results strongly suggest that deep architectures are necessary to model thekind of complexity exhibited in raw Web data such as images, natural language text, music, and video. Newsuccessful principles for learning such deep architectures have been proposed, which will be the starting pointof the proposed research. At the heart of these models as well as at the heart of many recent Web applicationsof machine learning is the automatic extraction of features in which Web objects (e.g., two Web pages, twosongs, two sentences, two sequences of images) can be compared, or which can form the basis of predictivesystems. Here we propose to focus on, improve, and apply learning algorithms for deep architectures that canconstruct a more meaningful representation for Web data with a sequential nature (such as text, music, andvideo). Such algorithms would be useful for a number of Web applications involving large quantities of suchdata, by allowing to obtain compact summarizations of such data, to provide the high-level representationsfrom which predictions and classifications are easier to make, and to meaningfully compare Web objects forrecommendation systems.
Programme
Strategic Projects Supplemental Competition
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
356722-2007
Name nom
Bengio, Yoshua
Department département
Informatique et recherche opérationnelle
Organization id
63
Institution établissement
Université de Montréal
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2007
Competition year année de concours
2007
Award amount
99000
Program id
STPSC
Program name en
Strategic Projects Supplemental Competition
Program name fr
Projets stratégiques compétition supplémentaire
Group en
Research partnerships programs
Group fr
Programmes de partenariat
Committee code
1064
Committee name en
Strategic Advanced Communications and Management of Information 1
Committee name fr
Stratégique -Technologies avancées de communication et de gestion de l'information 1
Area of application code
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
Research subject code
2805
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Learning and inference theories
Research subject fr
Théories de l'apprentissage et l'inférence
Partie
1
Nb partie
2
Application title
Deep learning of similarity spaces for sequential web data
Keyword
NA
Application summary
Recent theoretical and empirical work in statistical machine learning has demonstrated the importance ofdeveloping learning algorithms for deep architectures, i.e., function classes obtained by composing multiplenon-linear transformations. These results strongly suggest that deep architectures are necessary to model thekind of complexity exhibited in raw Web data such as images, natural language text, music, and video. Newsuccessful principles for learning such deep architectures have been proposed, which will be the starting pointof the proposed research. At the heart of these models as well as at the heart of many recent Web applicationsof machine learning is the automatic extraction of features in which Web objects (e.g., two Web pages, twosongs, two sentences, two sequences of images) can be compared, or which can form the basis of predictivesystems. Here we propose to focus on, improve, and apply learning algorithms for deep architectures that canconstruct a more meaningful representation for Web data with a sequential nature (such as text, music, andvideo). Such algorithms would be useful for a number of Web applications involving large quantities of suchdata, by allowing to obtain compact summarizations of such data, to provide the high-level representationsfrom which predictions and classifications are easier to make, and to meaningfully compare Web objects forrecommendation systems.