[Pycon] [new paper] "Francesco Farina" - Learning from Constraints: a Distributed and Privacy-Preserving Approach

[info a pycon.it]

Title: Learning from Constraints: a Distributed and Privacy-Preserving Approach
Duration: 60 (includes Q&A)
Q&A Session: 15
Language: en
Type: Talk

Abstract: Learning from Constraints (LFC) reframes the learning process in a context that is described by a collection of constraints. Such constraints are the mean that is used to inject knowledge into the learning process and they represent different aspects of the task at hand, like the knowledge on relationships among classes, on interactions among different tasks and on mutual exclusivity of predictions.

While LFC has always been conceived as a centralized framework, in this talk we present an extension of LFC to a distributed setting, where multiple computational nodes, connected over the network, contribute to the learning process. This setting is inspired by the nowadays organization of data and knowledge. It is extremely common to participate to communities, to share some resources (e.g., photos on social networks) and to keep other private (e.g., pictures saved on the cloud). Moreover, there is an increasing need of customized or more robust services that might benefit both from private and public data (e.g., a recognizer of pictures of a custom type).
The distributed implementation of LFC proposed in this talk is based on an asynchronous distributed optimization algorithm called ASYMM.

Some knowledge of optimization algorithms, Numpy, Networkx and Tensorflow is suggested, but not required.

Tags: [u'Machine Learning', u'numpy', u'tensorflow', u'distributed', u'Big-Data', u'Learning-from-Constraints', u'Algorithms']