Talks

So far artificial neural networks have been applied to discover phase diagrams in many different physical models. However, none of these studies have revealed any fundamentally new physics. A major problem is that these neural networks are mainly considered as black box algorithms. On the journey to detect new physics it is important to interpret what artificial neural networks learn. On the one hand this allows us to judge whether to trust the results, and on the other hand this can give us insight to possible new physics. In this talk I will 
discuss applications to different models where we successfully interpreted what was learned by the neural networks.

In the usual paradigm of quantum error correction, the information to be protected can be encoded in a system of abstract qubits or modes.  But how does this work for physical information, which cannot be described in this way?  Just as direction information cannot be conveyed using a sequence of words if the parties involved do not share a reference frame, physical quantum information cannot be conveyed using a sequence of qubits or modes without a shared reference frame.  Covariant quantum error correction is a procedure for protecting such physical information against noise in such a way that the encoding and decoding operations transform covariantly with respect to an external symmetry group. In this talk, we'll study covariant QEC, and we will see that there do not exist finite dimensional quantum codes that are covariant with respect to continuous symmetries.  Conversely, we'll see that there do exist finite codes for finite groups, and continuous variable (CV) codes for continuous groups.  This leads to a CV method of circumventing the Eastin-Knill theorem.  By relaxing our requirements to allow for only approximate error correction and covariance, we'll find a fundamental tension between a code's ability to approximately correct errors and covariance with respect to a continuous symmetry.  In this way, we'll arrive at an approximate version of the Eastin-Knill theorem, and we'll end by learning what covariant QEC tells us about continuous symmetries in AdS/CFT, among other applications. 

In some of the planet's most extreme environments scientists are constructing enormous detectors to study the very rare interactions produced by neutrinos. In particular, at South Pole Station Antarctica more than a cubic kilometer of the deep glacial ice has been instrumented to construct the world's largest neutrino detector to date: the IceCube Neutrino Observatory.  Designed to detect the highest energy neutrinos expected to be produced in astrophysical processes, IceCube has established a vibrant scientific program that has begun to revolutionize the fields of particle and astro-physics. In this talk I will present some of the most recent results from this new window to the Universe, and will discuss the plans underway to significantly enhance its long-term future reach. 

Epistemic interpretations of quantum theory maintain that quantum states only represent incomplete information about the physical states of the world. A major motivation for this view is the promise to provide a reasonable account of state update under measurement by asserting that it is simply a natural feature of updating incomplete statistical information. Here we demonstrate that all known epistemic ontological models of quantum theory in dimension d ≥ 3, including those designed to evade the conclusion of the PBR theorem, cannot represent state update correctly. Conversely, interpretations for which the wavefunction is real evade such restrictions despite remaining subject to long-standing criticism regarding physical discontinuity, indeterminism and the ambiguity of the Heisenberg cut. This revives the possibility of a no-go theorem with no additional assumptions, and demonstrates that what is usually thought of as a strength of epistemic interpretations may in fact be a weakness. We also discuss hidden Markov models and their relationship to ontological models, demarcating the ways in which one might move ‘outside’ the ontological models formalism.

The construction of soluble lattice toy models is an important theoretical approach in the study of strongly interacting topological phases of matter. On the other hand, the primary experimental probe to such systems is via electromagnetic response. Somewhat unsatisfactorily, the current systematic construction of the lattice toy models focuses on braiding statistics and does not admit coupling to an electromagnetic background. Thus there is a mismatch between our theoretical approach and experimental probe. In this talk I introduce how to systematically incorporate electromagnetic response into the soluble lattice toy models, for a large class of abelian topological phases. [Reference: 1902.06756] 

Despite intensive theoretical research for several decades, the theory of quantum gravity remains elusive. I will review the obstacles that prevent from reconciling the principles of general relativity with those of quantum mechanics. It is plausible that an eventual ultraviolet completion of general relativity will require sacrificing some of these principles. I will then focus on the class of theories where the abandoned property is local Lorentz invariance, replaced by an approximate anisotropic scaling symmetry in deep ultraviolet. At low energies these theories reduce to a special type of scalar — tensor gravity. I will show that this approach allows us to construct renormalizable gravitational theories in any number of spacetime dimensions. The study of a (2+1) dimensional model reveals its asymptotic freedom and suggests that this property may be generic for gravity with anisotropic scaling. Relevance of these results for gravity in the real world will be discussed.