Talks

In this talk, I will discuss some interesting connections between Hamiltonian complexity, error correction, and quantum circuits. First, motivated by the Quantum PCP Conjecture, I will describe a construction of a family of local Hamiltonians where the complexity of ground states — even when subject to large amounts of noise — is superpolynomial (under plausible complexity assumptions). The construction is simple, making use of the well-known Feynman-Kitaev circuit Hamiltonian construction.

Next, I will describe how this complexity result can be repurposed to construct local approximate error correcting codes with (nearly) linear distance and (nearly) constant rate. By contrast, it remains an open problem to construct quantum low-density parity check (QLDPC) stabilizer codes that achieve similarly good parameters.

An interesting component of our code construction is the Brueckmann-Terhal spacetime circuit Hamiltonian construction, and a novel analysis of its spectral gap. Our codes are ground spaces of non-commuting  — but frustration-free -- local Hamiltonians. We believe that our results present additional evidence that the study of approximate error correction and non-stabilizer codes may be much richer than previously thought.

This talk will have more questions than answers, and is based on two papers:

·  https://arxiv.org/abs/1802.07419 (joint with Chinmay Nirkhe and Umesh Vazirani)

·  https://arxiv.org/abs/1811.00277 (joint work with Thom Bohdanowicz, Elizabeth Crosson, and Chinmay Nirkhe)

Gravity theories naturally allow for edge states generated by non-trivial boundary-condition preserving diffeomorphisms. I present a specific set of boundary conditions inspired by near horizon physics, show that it leads to soft hair excitations of black hole solutions and discuss implications for black hole entropy.

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.