**TAMÁS KRIVÁCH**

**Université de Genève**

Computational limitations tend to appear in many areas when addressing quantum information questions on a network. In this talk I will introduce you to novel approaches, relying on machine learning techniques, to tackle problems in two commonly examined scenarios: NPA hierarchies and causal networks. For the NPA hierarchies SDP is in technically comuptationally efficient, however is limited when problem sizes become large. To circumvent this, we train a machine which, given a partially complete matrix, tries to give a filling such that the matrix is positive semidefinite. For causal networks, we examine the case of ’the triangle’, which is a simple network with three sources and three parties, each source distributing to two parties. Our goal is to train a machine to decide whether a given distribution can be reproduced with classical sources or not by trying to construct a strategy to do so. For both lines of work I will introduce the necessary machine learning basics, describe our approach and show some promising preliminary results.**Seminar, November 15, 2018, 15:00. ICFO’s Seminar Room
Hosted by Prof. Antonio Acín**