Search Talks

I’ll talk about two independent works on classical and quantum neural networks connected by information theory. In the first part of the talk, I’ll treat sequence models as one-dimensional classical statistical mechanical systems and analyze the scaling behavior of mutual information. I'll provide a new perspective on why recurrent neural networks are not good at natural language processing. In the second part of the talk, I’ll study information scrambling dynamics when quantum neural networks are trained by classical gradient descent algorithm. For many problems, this hybrid quantum-classical training process consists of two stages where information scrambles very differently in the network. 

I propose [1] to use the residual anyons of overscreened Kondo physics for quantum computation. A superconducting proximity gap of Δ<TK can be utilized to isolate the anyon from the continuum of excitations and stabilize the non-trivial fixed point. We use the dynamical large-N technique [2] and bosonization to show that the residual entropy survives in a superconductor and suggest a charge Kondo setup for isolating and detecting the Majorana fermion in the two-channel Kondo impurity.
I will then conjecture that topological defects in a multichannel Kondo lattice carry anyons. Motivated by this, we look at a two-channel SU(N) Kondo lattice in the large-N limit [3].  In this model, the continuous channel-symmetry is spontaneously broken, forming a “channel ferromagnet” and realizing the so-called fractionalized order parameter [4]. By integrating out the fermions we derive an effective action that describes the symmetry breaking and its emergent collective modes. Remarkably, topological defects in the order parameter carry a U(1) flux, manifested in the Aharonov-Bohm phase picked by electrons that orbit the defect. We argue that the phase diagram contains a non-magnetic transition between a large and a small Fermi surface.
I will also briefly highlight our recent results [5] on a magnetically frustrated Kondo-screened triangle which contains two symmetry-preserving phases, transcending the Landau-Ginzburg paradigm. The quantum phase transition is driven by the proliferation of instantons of the emergent gauge theory and can be regarded as a toy model for the deconfined criticality.

Motivated by puzzles in quantum gravity AdS/CFT, Lenny Susskind posed the following question: supposing one had the technological ability to distinguish a macroscopic superposition of two given states |v> and |w> from incoherent mixture of those states, would one also have the technological ability to map |v> to |w> and vice versa?  More precisely, how does the quantum circuit complexity of the one task relate to the quantum circuit complexity of the other?  Here we resolve Susskind's question -- showing that the two complexities are essentially identical, even for approximate versions of these tasks, with the one caveat that a unitary transformation that maps |v> to |w> and |w> to -|v> need not imply any distinguishing ability.  Informally, "if you had the ability to prove Schrödinger's cat was in superposition, you'd necessarily also have the ability to bring a dead cat back to life."  I'll also discuss the optimality of this little result and some of its implications.

Paper (with Yosi Atia) in preparation

Last fall, a team at Google announced the first-ever demonstration of "quantum computational supremacy"---that is, a clear quantum speedup over a classical computer for some task---using a 53-qubit programmable superconducting chip called Sycamore.  In addition to engineering, Google's accomplishment built on a decade of research in quantum computing theory.  This talk will discuss questions like: what exactly was the contrived computational problem that Google solved?  How does one verify the outputs using a classical computer?  And how confident are we that the problem really is classically hard---especially in light of subsequent counterclaims by IBM?  I'll end with a proposed application for Google's experiment---namely, the generation of certified random bits, for use (for example) in proof-of-stake cryptocurrencies---that I've been developing and that Google is now working to demonstrate.

We propose a unified manner of understanding two important phenomena: color confinement in large-N gauge theory, and Bose-Einstein condensation (BEC). We do this by clarifying the relation between the standard criteria, based on the off-diagonal long range order (ODLRO) for the BEC and the Polyakov loop for gauge theory: the constant offset of the distribution of Polyakov line phase is ODLRO. Indistinguishability associated with the symmetry group --- SU(N) or O(N) in gauge theory, and S_N permutation in the system of identical bosons --- plays the key role in both cases. This viewpoint may have implications to confinement at finite N, and to quantum gravity via gauge/gravity duality. 

Meta-learning involves learning mathematical devices using problem instances as training data. In this talk, we first describe recent meta-learning approaches involving the learning of objects such as: initial weights, parameterized losses, hyper-parameter search strategies, and samplers. We then discuss learned optimizers in further detail and their applications towards optimizing variational circuits. This talk also covers some lessons learned starting a spin-off from academia.

We propose a model for combining the Standard Model (SM) with gravity. It relies on a non-minimal coupling of the Higgs field to the Ricci scalar and on the Palatini formulation of gravity. Without introducing any new degrees of freedom in addition to those of the SM and the graviton, this scenario achieves two goals. First, it generates the electroweak symmetry breaking by a non-perturbative gravitational effect. In this way, it does not only address the hierarchy problem but opens up the possibility to calculate the Higgs mass. Second, the model incorporates inflation at energies below the onset of strong-coupling of the theory. Provided that corrections due to new physics above the scale of inflation are not unnaturally large, we can relate inflationary parameters to data from collider experiments.
 

Cosmologists wish to explain how our universe, in all its complexity, could ever have come about. This is the problem of initial conditions and the first step towards its solution is the assessment of the universe’s entropy today. It is widely agreed upon that the entropy of vacuum energy, given by the Bekenstein bound, makes up the bulk of the current entropy budget, dominating over that of gravity and over thermal motions of the cosmic radiation background.

Have we counted all significant contributions to the entropy inventory? There is one number which we have not considered: the number of degrees of freedom in our vibrant, complex biosphere. What is the entropy of life? Is it sizeable enough to need to be accounted for at the Big Bang, or negligible compared to vacuum entropy?