Q-Math Seminar

Logo Q-Math

Florio M. Ciaglia (UC3M)

Can Cencov meet Petz?

Friday the 27th of January, 2023, 13:00, Room 2.2.D08

Cencov's theorem on the uniqueness of the Fisher-Rao metric tensor and Petz's classification of monotone metric tensors are considered the cornerstones of classical and quantum information geometry, respectively. I will review and discuss these beautiful results with particular attention to a categorical formalization of Petz's classification in analogy with Cencov's ideology, and I will then argue about a possible unification in the framework of Jordan algebras.

Link for online session (Active on request):

Satoya Imai (University of Siegen)

Geometry of quantum entangled states

Friday the 25th of November, 2022, 13:00, Room 2.2.D08

In this talk, I will first give an introduction of quantum entanglement in the language of quantum information. Then I will explain the problem of whether a given state is entangled or not. Finally I will provide several methods to detect entanglement from geometrical viewpoints [1-4].

[1] S Imai, N Wyderka, A Ketterer, O Gühne, Physical Review Letters 126 (15), 150501
[2] XD Yu, S Imai, O Gühne, Physical Review Letters 127 (6), 060504
[3] A Ketterer, S Imai, N Wyderka, O Gühne, Physical Review A 106 (1), L010402
[4] some ongoing projects.

Link for online session:

Alejandro Pozas-Kerstjens (ICMAT)

Quantum-inspired solutions for privacy leaks in machine learning

Friday the 11th of November, 2022, 13:00, Room 2.2.D08

Vast amounts of data are routinely processed in machine learning pipelines, every time covering more aspects of our interactions with the world. However, the quest for performance is leaving other important aspects, such as privacy, on the side. For example, when the models processing the data are made public, is the safety of the data used for training it guaranteed? This is a question of utmost importance especially when processing sensitive data such as medical records.
In this talk, I will argue and practically illustrate that insights in quantum information, concretely coming from the tensor network representations of quantum many-body states, can help in devising better privacy-preserving machine learning algorithms. In the first part, I will show that standard neural networks are vulnerable to a type of privacy leak that involves global properties of the data used for training, thus being a priori resistant to standard protection mechanisms. In the second, I will show that tensor networks, when used as machine learning architectures, are invulnerable to this vulnerability. The proof of the resilience is based on the existence of canonical forms for tensor networks. Given the growing expertise in training tensor networks and the recent interest in tensor-based reformulations of popular machine learning architectures, these results imply that one may not have to be forced to make a choice between accuracy in prediction and ensuring the privacy of the information processed when using machine learning on sensitive data.

Nicolai Rothe (Bergische Universität Wuppertal)

De Sitter Solutions to the Semiclassical Einstein Equation in Cosmology

Friday the 21st of October, 2022, 13:00, Room 2.2.D08

We present a characterization of cosmological de Sitter solutions to the semiclassical Einstein equation (SCE) with a free scalar quantum field in the Bunch-Davies state. In this setting, the SCE may be viewed as a (non-dynamic) consistency equation for the parameters of the model. Our approach allows to identify parameter settings in which there exist multiple cosmological de Sitter solutions with expansion rates $H$ differing by several (in fact, arbitrarily many) magnitudes for only one field theory. By these observations a quantum field is, in principle, capable to drive both an inflationary phase (approximated by the large-$H$ solution) and a dark energy-dominated phase (approximated by the small-$H$ solution) in the expansion of the universe.

Link for online session:

Julio de Vicente (UC3M)

Genuine multipartite entanglement of quantum states in the multiple-copy scenario

Friday the 14th of October, 2022, 13:00, Room 2.2.D08

Genuine multipartite entanglement (GME) is considered a powerful form of entanglement since it corresponds to those states that are not biseparable, i.e. a mixture of partially separable states across different bipartitions of the parties. In this talk I will study this phenomenon in the multiple-copy regime, where many perfect copies of a given state can be produced and controlled. In this scenario the above definition leads to subtle intricacies as biseparable states can be GME-activatable, i.e. several copies of a biseparable state can display GME. I will show, however, that the set of GME-activatable states admits a simple characterization: a state is GME-activatable if and only if it is not partially separable across one bipartition of the parties. This leads to the second question of whether there is a general upper bound in the number of copies that needs to be considered in order to observe the activation of GME, which will be answered in the negative. In particular, by providing an explicit construction, it can be proven that for any number of parties and any number $k\in\mathbb{N}$ there exist GME-activatable multipartite states of fixed (i.e. independent of $k$) local dimensions such that $k$ copies of them remain biseparable. This is joint work with Carlos Palazuelos.

Zhen-Peng Xu (Universität Siegen)

Quantum networks cannot generate graph states with high fidelity

Friday the 23rd of September, 2022, 13:00, Room 2.2.D08

Quantum networks lead to novel notions of locality and correlations which are not well understood. An important problem concerns the question which states can be experimentally prepared with a given network structure and which not. By exploiting the inflation technique and symmetry analysis, we prove that all multi-qubit graph states arising from a connected graph cannot originate from any bipartite network. Moreover, the fidelity of a multi-qubit graph state and any network state cannot exceed $9/10$. Similar results can also be established for a large class of multi-qudit graph states. More specifically, the fidelity of any dimensional graph state with constant multiplicities and states prepared in a bipartite network cannot exceed $0.95495$.