arXiv subjects

Bradd Hart

Publications and source records attributed to Bradd Hart.

At least 19 recordsLinked to original sources

arXiv2026

Totally Bounded Elements in W*-probability Spaces

We introduce the notion of a totally ($K$-) bounded element of a $W^*$-probability space $(M, φ)$ and, borrowing ideas of Kadison, give an intrinsic characterization of the $^*$-subalgebra $M_{\operatorname{tb}}$ of totally bounded elements. Namely, we show that $M_{\operatorname{tb}}$ is the unique strongly dense $^*$-subalgebra $M_0$ of totally bounded elements of $M$ for which the collection of totally $1$-bounded elements of $M_0$ is complete with respect to the $\|\cdot\|_φ^\#$-norm and for which $M_0$ is closed under all operators $h_a(\log(Δ))$ for $a \in \mathbb{N}$, where $Δ$ is the modular operator and $h_a(t):=1/\cosh(t-a)$ (see Theorem 4.3). We also prove that totally $K$-bounded elements of an Ocneanu ultraproduct admit representatives with the same total bound using a careful effective estimate of the distance of a given totally bounded element to the totally 1-bounded elements. An alternative proof in the appendix uses an isometric $H^\infty$-lifting theorem for the Ocneanu multiplier quotient, derived from a metric $H^\infty$-lifting theorem for $C^*$-quotients and a $C^*$-algebraic Schur parametrization. We combine these results with Rieffel and Van Daele's bounded operator approach to modular theory to arrive at a new language and axiomatization of $W^*$-probability spaces as metric structures. Previous work of Dabrowski had axiomatized $W^*$-probability spaces using a smeared version of multiplication, but the subalgebra $M_{\operatorname{tb}}$ allows us to give an axiomatization in terms of the original algebra operations. Finally, we prove the (non-)axiomatizability of several classes of $W^*$-probability spaces.

math.OA↗

arXiv2025

Quantifying The Limits of AI Reasoning: Systematic Neural Network Representations of Algorithms

A main open question in contemporary AI research is quantifying the forms of reasoning neural networks can perform when perfectly trained. This paper answers this by interpreting reasoning tasks as circuit emulation, where the gates define the type of reasoning; e.g. Boolean gates for predicate logic, tropical circuits for dynamic programming, arithmetic and analytic gates for symbolic mathematical representation, and hybrids thereof for deeper reasoning; e.g. higher-order logic. We present a systematic meta-algorithm that converts essentially any circuit into a feedforward neural network (NN) with ReLU activations by iteratively replacing each gate with a canonical ReLU MLP emulator. We show that, on any digital computer, our construction emulates the circuit exactly--no approximation, no rounding, modular overflow included--demonstrating that no reasoning task lies beyond the reach of neural networks. The number of neurons in the resulting network (parametric complexity) scales with the circuit's complexity, and the network's computational graph (structure) mirrors that of the emulated circuit. This formalizes the folklore that NNs networks trade algorithmic run-time (circuit runtime) for space complexity (number of neurons). We derive a range of applications of our main result, from emulating shortest-path algorithms on graphs with cubic--size NNs, to simulating stopped Turing machines with roughly quadratically--large NNs, and even the emulation of randomized Boolean circuits. Lastly, we demonstrate that our result is strictly more powerful than a classical universal approximation theorem: any universal function approximator can be encoded as a circuit and directly emulated by a NN.

cs.LG↗

arXiv2024

Non-locally modular regular types in classifiable theories

We introduce the notion of strong $p$-semi-regularity and show that if $p$ is a regular type which is not locally modular then any $p$-semi-regular type is strongly $p$-semi-regular. Moreover, for any such $p$-semi-regular type, "domination implies isolation" which allows us to prove the following: Suppose that $T$ is countable, classifiable and $M$ is any model. If $p\in S(M)$ is regular but not locally modular and $b$ is any realization of $p$ then every model $N$ containing $M$ that is dominated by $b$ over $M$ is both constructible and minimal over $Mb$.

math.LO↗

arXiv2023

The undecidability of having the QWEP

We show that neither the class of C*-algebras with Kirchberg's QWEP property nor the class of W*-probability spaces with the QWEP property are effectively axiomatizable (in the appropriate languages). The latter result follows from a more general result, namely that the hyperfinite III$_1$ factor does not have a computable universal theory in the language of W*-probability spaces. We also prove that the Powers' factors $\mathcal{R}_λ$, for $0<λ<1$, when equipped with their canonical Powers' states, do not have computable universal theory.

math.OA↗

arXiv2023

An Introduction To Continuous Model Theory

We present an introduction to modern continuous model theory with an emphasis on its interactions with topics covered in this volume such as $C^*$-algebras and von Neumann algebras. The role of ultraproducts is highlighted and expositions of definable sets, imaginaries, quantifier elimination and separable categoricity are included.

math.OA↗

arXiv2023

Locally universal C*-algebras with computable presentations

The Kirchberg Embedding Problem (KEP) asks if every C*-algebra embeds into an ultrapower of the Cuntz algebra $\mathcal{O}_2$. In an effort to provide a negative solution to the KEP and motivated by the recent refutation of the Connes Embedding Problem, we establish two computability-theoretic consequences of a positive solution to KEP. Both of our results follow from the a priori weaker assumption that there exists a locally universal C*-algebra with a computable presentation.

math.LO↗

arXiv2023

A quantum complexity approach to the Kirchberg Embedding Problem

The Kirchberg Embedding Problem (KEP) asks if every C*-algebra embeds into an ultrapower of the Cuntz algebra $\cal O_2$. Motivated by the recent refutation of the Connes Embedding Problem using the quantum complexity result MIP*=RE, we establish two quantum complexity consequences of a positive solution to KEP. Both results involve almost-commuting strategies to nonlocal games.

math.OA↗

arXiv2022

On Tsirelson pairs of C*-algebras

We introduce the notion of a Tsirelson pair of C*-algebras, which is a pair of C*-algebras for which the space of quantum strategies obtained by using states on the minimal tensor product of the pair and the space of quantum strategies obtained by using states on the maximal tensor product of the pair coincide. We exhibit a number of examples of such pairs that are "nontrivial" in the sense that the minimal tensor product and the maximal tensor product of the pair are not isomorphic. For example, we prove that any pair containing a C*-algebra with Kirchberg's QWEP property is a Tsirelson pair. We then introduce the notion of a C*-algebra with the Tsirelson property (TP) and establish a number of closure properties for this class. We also show that the class of C*-algebras with the TP form an axiomatizable class (in the sense of model theory), but that this class admits no "effective" axiomatization.

math.OA↗

arXiv2022

A survey on the model theory of tracial von Neumann algebras

We survey the developments in the model theory of tracial von Neumann algebras that have taken place in the last fifteen years. We discuss the appropriate first-order language for axiomatizing this class as well as the subclass of II$_1$ factors. We discuss how model-theoretic ideas were used to settle a variety of questions around isomorphism of ultrapowers of tracial von Neumann algebras with respect to different ultrafilters before moving on to more model-theoretic concerns, such as theories of II$_1$ factors and existentially closed II$_1$ factors. We conclude with two recent applications of model-theoretic ideas to questions around relative commutants.

math.LO↗

arXiv2022

Approximate traces on groups and the quantum complexity class $\operatorname{MIP}^{co,s}$

An open question in quantum complexity theory is whether or not the class $\operatorname{MIP}^{co}$, consisting of languages that can be efficiently verified using interacting provers sharing quantum resources according to the quantum commuting model, coincides with the class $coRE$ of languages with recursively enumerable complement. We introduce the notion of a qc-modulus, which encodes approximations to quantum commuting correlations, and show that the existence of a computable qc-modulus gives a negative answer to a natural variant of the aforementioned question.

cs.CC↗

arXiv2022

The almost sure theory of finite metric spaces

We establish an approximate zero-one law for sentences of continuous logic over finite metric spaces of diameter at most $1$. More precisely, we axiomatize a complete metric theory $T_{\mathrm{as}}$ such that, given any sentence $σ$ in the language of pure metric spaces and any $ε>0$, the probability that the difference of the value of $σ$ in a random metric space of size $n$ and the value of $σ$ in any model of $T_{\mathrm{as}}$ is less than $ε$ approaches $1$ as $n$ approaches infinity. We also establish some model-theoretic properties of the theory $T_{\mathrm{as}}$.

math.LO↗

arXiv2021

The Universal Theory Of The Hyperfinite II$_1$ Factor Is Not Computable

We show that the universal theory of the hyperfinite II$_1$ factor is not computable. The proof uses the recent result that MIP*=RE. Combined with an earlier observation of the authors, this yields a proof that the Connes Embedding Problem has a negative solution that avoids the equivalences with Kirchberg's QWEP Conjecture and Tsirelson's Problem.+

math.LO↗

arXiv2020

Properties expressible in small fragments of the theory of the hyperfinite II_1 factor

We show that any II$_1$ factor that has the same 4-quantifier theory as the hyperfinite II$_1$ factor $\mathcal{R}$ satisfies the conclusion of the Popa Factorial Commutant Embedding Problem (FCEP) and has the Brown property. These results improve recent results proving the same conclusions under the stronger assumption that the factor is actually elementarily equivalent to $\mathcal{R}$. In the same spirit, we improve a recent result of the first-named author, who showed that if (1) the amalgamated free product of embeddable factors over a property (T) base is once again embeddable, and (2) $\mathcal{R}$ is an infinitely generic embeddable factor, then the FCEP is true of all property (T) factors. In this paper, it is shown that item (2) can be weakened to assume that $\mathcal{R}$ has the same 3-quantifier theory as an infinitely generic embeddable factor.

math.OA↗

arXiv2020

Operator algebras with hyperarithmetic theory

We show that the following operator algebras have hyperarithmetic theory: the hyperfinite II$_1$ factor $\mathcal R$, $L(Γ)$ for $Γ$ a finitely generated group with solvable word problem, $C^*(Γ)$ for $Γ$ a finitely presented group, $C^*_λ(Γ)$ for $Γ$ a finitely generated group with solvable word problem, $C(2^ω)$, and $C(\mathbb P)$ (where $\mathbb P$ is the pseudoarc). We also show that the Cuntz algebra $\mathcal O_2$ has a hyperarithmetic theory provided that the Kirchberg embedding problem has an affirmative answer. Finally, we prove that if there is an existentially closed (e.c.) II$_1$ factor (resp. C$^*$-algebra) that does not have hyperarithmetic theory, then there are continuum many theories of e.c. II$_1$ factors (resp. e.c. C$^*$-algebras).

math.OA↗

arXiv2019

Correspondences, Ultraproducts and Model Theory

We study correspondences of tracial von Neumann algebras from the model-theoretic point of view. We introduce and study an ultraproduct of correspondences and use this ultraproduct to prove, for a fixed pair of tracial von Neumann algebras M and N, that the class of M-N correspondences forms an elementary class. We prove that the corresponding theory is classifiable, all of its completions are stable, that these completions have quantifier elimination in an appropriate language, and that one of these completions is in fact the model companion. We also show that the class of triples (M, H, N), where M and N are tracial von Neumann algebras and H is an M-N correspondence, form an elementary class. As an application of our framework, we show that a II_1 factor M has property (T) precisely when the set of central vectors form a definable set relative to the theory of M-M correspondences. We then use our approach to give a simpler proof that the class of structures (M, Phi), where M is a sigma-finite von Neumann algebra and Phi is a faithful normal state, forms an elementary class. Finally, we initiate the study of a family of Connes-type ultraproducts on C*-algebras.

math.LO↗

arXiv2019

The Urysohn sphere is pseudofinite

We show that the Urysohn sphere is pseudofinite. As a consequence, we derive an approximate $0$-$1$ law for finite metric spaces of diameter at most $1$.

math.LO↗

arXiv2018

Universal covers of commutative finite Morley rank groups

We give an algebraic description of the structure of the analytic universal cover of a complex abelian variety which suffices to determine the structure up to isomorphism. More generally, we classify the models of theories of "universal covers" of rigid divisible commutative finite Morley rank groups.

math.LO↗

arXiv2017

Explicit sentences distinguishing McDuff's II$_1$ factors

Recently, Boutonnet, Chifan, and Ioana proved that McDuff's examples of continuum many pairwise non-isomorphic separable II$_1$ factors are in fact pairwise non-elementarily equivalent. Their proof proceeded by showing that any ultrapowers of any two distinct McDuff examples are not isomorphic. In a paper by the first two authors of this paper, Ehrenfeucht-Fraïsse games were used to find an upper bound on the quantifier complexity of sentences distinguishing the McDuff examples, leaving it as an open question to find concrete sentences distinguishing the McDuff factors. In this paper, we answer this question by providing such concrete sentences.

math.LO↗