25.4.2018 11.15 - 12.00 XVII,
(University of Toronto)
Machine Learning and Quantum Physics: Setting Realistic Expectations
The growing intersection between machine learning and quantum physics is beginning to repeat patterns familiar from classical machine learning: the gap between theoretical research and the practice is widening. On the theory side, we have outstanding results on generalizing statistical learning theory, we discovered and exploited connections between the representation power of deep learning networks and the methods used in many-body physics, and we have many quantum learning protocols with a promised exponential advantage. On the practical side, we have a handful of basic quantum optimization and sampling routines running on small-scale quantum computers and elementary feedforward architectures can recover dusty results in the study of phase transitions. This talk will give an introduction to the topic and highlight specific problems that are worth addressing to move the field forward.
27.4.2018 13.15 - 14.00 XVII,
Reinhard Werner (Leibniz Universität Hannover)
Uncertainty relations for measurement
The well-known textbook uncertainty relation makes a statement about the impossibility to prepare states, which have sharp distributions for both position and momentum.
This is only one aspect of Heisenberg's original work, and completely leaves out the issue he discusses with his famous gamma-ray microscope: The disturbance of momentum by an approximate position measurement. I will discuss how to formulate this aspect in a quantitative way as the impossibility of realizing accurate joint measurements of position and momentum. Curiously, the resulting uncertainty relations are formally exactly the same as for preparation, but with a quite different interpretation of Delta P and Delta Q. It turns out that this is a coincidence due to the high symmetry of such Fourier-related pairs. The setup presented works for any pair (or n-tuple) of observables, and in this setting such equality generally fails, although there are some inequalities.