Mixtures of Three-way Models for Multi-relational Data.
Supervision : Antoine Bordes, CNRS - Université de Technologie de Compiègne.
Dates : position open from January 1st, 2014. (earlier or later start dates can be negotiable)
Project description :
Many data such as Knowledge Bases (e.g. Freebase) are multi-relational, in that they describe multiple relations between entities. While there is a large body of work focused on modeling these data, modeling these multiple types of relations jointly remains challenging. This research project targets data with large numbers of relation types (more than 10k), and for which the various relation types have heterogeneous properties like different connectivities or occurrence frequencies for instance. We propose to take this into account by using different relational three-way models (e.g. variants of tensor factorization models, with different loss functions, architectures, constraints, etc.) for different relation types. These models would be trained jointly and share some parameters (e.g. those encoding the entities), leading to an overall mixture of three-way models.
Such models will be based on previous work by Bordes et al. on modeling multi-relational data (see this publication page for recent papers) and will be developed in collaboration with Google.
A post-doctoral position is available as part of a Google Reseach Award obtained by Antoine Bordes. Research will be conducted within the French ANR project EVEREST on “lEarning high-leVEl REpresentations of large Sparse Tensors” being undertaken by Heudiasyc laboratory in Université de Technologie de Compiègne, with a partnership of Xerox Research Center Europe (Grenoble, France). See overview for more details on the project.
The post-doctoral fellow will be based in the Heudiasyc laboratory in Compiègne (France) and join the DI team headed by Yves Grandvalet. He/she will be supervised by Antoine Bordes. Heudiasyc is a joint laboratory with the Université de Technologie de Compiègne (UTC) and the French governmental agency for research (CNRS). In 2011, it was rated A+ (the highest rate) by the French Research evaluation agency (AERES). Heudiasyc fosters interdisciplinary research on information science and technology including machine learning, uncertain reasoning, operations research, robotics and knowledge management. In 2011 Heudiasyc was awarded with an excellence project (LabEx) on the « Control of Technological Systems of Systems ».
The fellowship is funded through a Google Research Award and will start after January 1st, 2014 for one year (currently 2500€ per month – gross salary).
The candidate should have a PhD or equivalent in computer science or mathematics. The following qualities are desirable : strong interests in machine learning, statistics or natural language processing; excellent record of academic and/or professional achievement; strong mathematical skills; strong programming skills ; good written and spoken communication skills in French or English. The ideal candidate should be able to conduct theoretical research, but also implement and test models on very large datasets.
Contact and application :
Applicants should send (preferably as a single PDF file):
Applications and inquiries should be directed to: Antoine Bordes - firstname.lastname@example.org