Mathematics (Faculty of)
http://hdl.handle.net/10012/9924
Tue, 16 Jul 2019 06:28:14 GMT2019-07-16T06:28:14ZCorrelation and Communication via a Quantum Field
http://hdl.handle.net/10012/14803
Correlation and Communication via a Quantum Field
Simidzija, Petar
We study the ability of qubit detectors to i) extract correlations from, and ii) transmit quantum information through, a quantum field.
We start by perturbatively studying the harvesting of correlations from thermal and squeezed coherent field states. We find that an increase in field temperature is detrimental to entanglement harvesting, but beneficial to mutual information harvesting. We also show that entanglement harvesting is independent of the field's coherent amplitude - which we relate to fundamental results regarding the entanglement structure of coherent field states - but strongly dependent on the field's squeezing amplitude. We conclude by analyzing the practical feasibility of entangling qubits using squeezed field states.
We then go on to study, non-perturbatively, the entanglement extraction by targets A and B from a quantum source S. After proving a general no-go theorem which applies for any A, B and S, we apply this theorem to the entanglement harvesting setup to prove that a wide class of i) degenerate, or ii) point-in-time coupled, detectors cannot harvest entanglement from any field state. We also discuss the role of communication in the process of entanglement extraction, and we end the chapter by presenting the simplest successful example of a non-perturbative entanglement harvesting protocol.
We conclude by studying the ability of flat spacetime observers Alice and Bob to transmit quantum information through a quantum field. We construct a perfect, field-mediated quantum channel, each use of which allows Alice to transmit a full qubit of information to Bob. This construction provides us with an understanding of how quantum information propagates through a relativistic field, which we find to be consistent with our understanding of the strong Huygens principle. Lastly, we analyze the possibility of simultaneously broadcasting a quantum message through a quantum field to multiple receivers, and discover severe fundamental limitations to such a setup.
Mon, 15 Jul 2019 00:00:00 GMThttp://hdl.handle.net/10012/148032019-07-15T00:00:00ZComputability Theory and Some Applications
http://hdl.handle.net/10012/14800
Computability Theory and Some Applications
Deveau, Michael
We explore various areas of computability theory, ranging from applications in computable structure theory primarily focused on problems about computing isomorphisms, to a number of new results regarding the degree-theoretic notion of the bounded Turing hierarchy.
In Chapter 2 (joint with Csima, Harrison-Trainor, Mahmoud), the set of degrees that are computably enumerable in and above $\mathbf{0}^{(\alpha)}$ are shown to be degrees of categoricity of a structure, where $\alpha$ is a computable limit ordinal. We construct such structures in a particularly useful way: by restricting the construction to a particular case (the limit ordinal $\omega$) and proving some additional facts about the widgets that make up the structure, we are able to produce a computable prime model with a degree of categoricity as high as is possible. This then shows that a particular upper bound on such degrees is exact.
In Chapter 3 (joint with Csima and Stephenson), a common trick in computable structure theory as it relates to degrees of categoricity is explored. In this trick, the degree of an isomorphism between computable copies of a rigid structure is often able to be witnessed by the clever choice of a computable set whose image or preimage through the isomorphism actually attains the degree of the isomorphism itself. We construct a pair of computable copies of $(\omega, <)$ where this trick will not work, examine some problems with decidability of the structures and work with $(\omega^2, <)$ to resolve them by proving a similar result.
In Chapter 4, the effectivization of Walker's Cancellation Theorem in group theory is discussed in the context of uniformity. That is, if we have an indexed collection of instances of sums of finitely generated abelian groups $A_i \join G_i \cong A_i \join H_i$ and the code for the isomorphism between them, then we wish to know to what extent we can give a single procedure that, given an index $i$, produces an isomorphism between $G_i$ and $H_i$.
Finally, in Chapter 5, several results pertaining to the bounded Turing degrees (also known as the weak truth-table degrees) and the bounded jump are investigated, with an eye toward jump inversion. We first resolve a potential ambiguity in the definition of sets used to characterize degrees in the bounded Turing hierarchy. Then we investigate some open problems related to lowness and highness as it appears in this realm, and then generalize a characterization about reductions to iterated bounded jumps of arbitrary sets. We use this result to prove the non-triviality of the hierarchy of successive applications of the bounded jump above any set, showing that the problem of jump inversion must be non-trivial if it is true in any relativized generality.
Mon, 15 Jul 2019 00:00:00 GMThttp://hdl.handle.net/10012/148002019-07-15T00:00:00ZApplications of Stochastic Gradient Descent to Nonnegative Matrix Factorization
http://hdl.handle.net/10012/14798
Applications of Stochastic Gradient Descent to Nonnegative Matrix Factorization
Slavin, Matthew
We consider the application of stochastic gradient descent (SGD) to the nonnegative matrix factorization (NMF) problem and the unconstrained low-rank matrix factorization problem. While the literature on the SGD algorithm is rich, the application of this specific algorithm to the field of matrix factorization problems is an unexplored area. We develop a series of results for the unconstrained problem, beginning with an analysis of standard gradient descent with a known zero-loss solution, and culminating with results for SGD in the general case where no zero-loss solution is assumed. We show that, with initialization close to a minimizer, there exist linear rate convergence guarantees.
We explore these results further with numerical experiments, and examine how the matrix factorization solutions found by SGD can be used as machine learning classifiers in two specific applications. In the first application, handwritten digit recognition, we show that our approach produces classification performance competitive with existing matrix factorization algorithms. In the second application, document topic classification, we examine how well SGD can recover an unknown words-to-topics matrix when the topics-to-document matrix is generated using the Latent Dirichlet Allocation model. This approach allows us to simulate two regimes for SGD: a fixed-sample regime where a large set of data is iterated over to train the model, and a generated-sample regime where a new data point is generated at each training iteration. In both regimes, we show that SGD can be an effective tool for recovering the hidden words-to-topic matrix. We conclude with some suggestions for further expansion of this work.
Mon, 15 Jul 2019 00:00:00 GMThttp://hdl.handle.net/10012/147982019-07-15T00:00:00ZUnitary Correlation Sets and their Applications
http://hdl.handle.net/10012/14793
Unitary Correlation Sets and their Applications
Harris, Samuel
We relate Connes' embedding problem in operator algebras to the Brown algebra Unc(n), which is defined as the universal unital C*-algebra generated by the entries of an n x n unitary matrix. In particular, we show that the embedding problem is equivalent to determining whether or not the algebra Unc(n) has the weak expectation property for any (equivalently, all) n at least 2. From this perspective, we develop a theory of what we call unitary correlation sets. These sets are analogous to the usual sets of bipartite probabilistic correlations arising from the typical models in quantum information theory. We show that the analogue of the weak Tsirelson problem for unitary correlation sets is again equivalent to Connes' embedding problem. Moreover, we show that as long as Alice and Bob's unitaries are of size at least 2, the set of spatial unitary correlations is never a closed set. This result is analogous to a recent theorem of Slofstra, which states that the set of quantum probabilistic correlations is not closed, so long as the input and output sets are large enough.
We also discuss several applications of the theory of unitary correlation sets. First, we show that the class of extended non-local games known as quantum XOR games is a rich enough class to detect the validity of Connes' embedding problem. That is, the embedding problem has a positive answer if and only if the value of every quantum XOR game in the commuting model agrees with the value of the game in the approximate finite-dimensional model. Second, we use a C*-algebraic analogue of the quantum teleportation and super-dense coding maps from quantum information theory to obtain separations between the tensor product model (or "quantum spatial" model) and the approximate finite-dimensional model (or "quantum approximate" model), for matrix-valued generalizations of the usual Tsirelson corrleation sets. We use some of the intermediate results to also obtain separations between the matrix versions of the finite-dimensional model (or "quantum model") and the tensor product model.
Thu, 04 Jul 2019 00:00:00 GMThttp://hdl.handle.net/10012/147932019-07-04T00:00:00Z