Bypassing the Monster: A Faster and Simpler Optimal Algorithm for Contextual Bandits Under Realizability

We consider the general (stochastic) contextual bandit problem under the realizability assumption, that is, the expected reward, as a function of contexts and actions, belongs to a general function class [Formula: see text]. We design a fast and simple algorithm that achieves the statistically optimal regret with only [Formula: see text] calls to an offline regression oracle across all T rounds. The number of oracle calls can be further reduced to [Formula: see text] if T is known in advance. Our results provide the first universal and optimal reduction from contextual bandits to offline regression, solving an important open problem in the contextual bandit literature. A direct consequence of our results is that any advances in offline regression immediately translate to contextual bandits, statistically and computationally. This leads to faster algorithms and improved regret guarantees for broader classes of contextual bandit problems.

Gaudin models and multipoint cnformal blocks: general theory

The construction of conformal blocks for the analysis of multipoint correlation functions with N > 4 local field insertions is an important open problem in higher dimensional conformal field theory. This is the first in a series of papers in which we address this challenge, following and extending our short announcement in [1]. According to Dolan and Osborn, conformal blocks can be determined from the set of differential eigenvalue equations that they satisfy. We construct a complete set of commuting differential operators that characterize multipoint conformal blocks for any number N of points in any dimension and for any choice of OPE channel through the relation with Gaudin integrable models we uncovered in [1]. For 5-point conformal blocks, there exist five such operators which are worked out smoothly in the dimension d.

On generators of the unstable $\mathscr A$-module $H^{*}((K(\mathbb F_2, 1))^{\times 5})$ in a generic degree and applications

Let us consider the prime field of two elements, $\mathbb F_2.$ It is well-known that the classical "hit problem" for a module over the mod 2 Steenrod algebra $\mathscr A$ is an interesting and important open problem of Algebraic topology, which asks a minimal set of generators for the polynomial algebra $\mathcal P_m:=\mathbb F_2[x_1, x_2, \ldots, x_m]$, regarded as a connected unstable $\mathscr A$-module on $m$ variables $x_1, \ldots, x_m,$ each of degree 1. The algebra $\mathcal P_m$ is the $\mathbb F_2$-cohomology of the product of $m$ copies of the Eilenberg-MacLan complex $K(\mathbb F_2, 1).$ Although the hit problem has been thoroughly studied for more than 3 decades, solving it remains a mystery for $m\geq 5.$ The aim of this work is of studying the hit problem of five variables. More precisely, we develop our previous work \cite{D.P3} on the hit problem for $\mathscr A$-module $\mathcal P_5$ in a degree of the generic form $n_t:=5(2^t-1) + 18.2^t,$ for any non-negative integer $t.$ An efficient approach to solve this problem had been presented. Moreover, we provide an algorithm in MAGMA for verifying the results and studying the hit problem in general. As an consequence, the calculations confirmed Sum's conjecture \cite{N.S2} for the relationship between the minimal sets of $\mathscr A$-generators of the polynomial algebras $\mathcal P_{m-1}$ and $\mathcal P_{m}$ in the case $m=5$ and degree $n_t.$ Two applications of this study are to determine the dimension of $\mathcal P_6$ in the generic degree $5(2^{t+4}-1) + n_1.2^{t+4}$ for all $t > 0$ and describe the modular representations of the general linear group of rank 5 over $\mathbb F_2.$ As a corollary, the cohomological "transfer", defined by W. Singer \cite{W.S1}, is an isomorphism at the bidegree $(5, 5+n_0).$ Singer's transfer is one of the relatively efficient tools to approach the structure of mod-2 cohomology of the Steenrod algebra.

Open Problems

Open problems in the study of topological 4-manifolds are explained in detail. An important open problem is to determine whether the disc embedding theorem and its antecedents hold for all groups; in other words, whether all groups are good. The disc embedding conjecture and the surgery conjecture are stated. The relationships between these conjectures and their various reformulations are explained. Of particular interest are the reformulations in terms of freely slicing certain infinite families of links. In particular, the surgery conjecture is true if and only if all good boundary links are freely slice. Good boundary links are the many-component analogues of Alexander polynomial one knots.

Circulant matrices with orthogonal rows and off-diagonal entries of absolute value 1

It is known that a real symmetric circulant matrix with diagonal entries d ≥ 0, off-diagonal entries ±1 and orthogonal rows exists only of order 2d + 2 (and trivially of order 1) [Turek and Goyeneche 2019]. In this paper we consider a complex Hermitian analogy of those matrices. That is, we study the existence and construction of Hermitian circulant matrices having orthogonal rows, diagonal entries d ≥ 0 and any complex entries of absolute value 1 off the diagonal. As a particular case, we consider matrices whose off-diagonal entries are 4th roots of unity; we prove that the order of any such matrix with d different from an odd integer is n = 2d + 2. We also discuss a similar problem for symmetric circulant matrices defined over finite rings ℤ m . As an application of our results, we show a close connection to mutually unbiased bases, an important open problem in quantum information theory.

Stabilizer extent is not multiplicative

The Gottesman-Knill theorem states that a Clifford circuit acting on stabilizer states can be simulated efficiently on a classical computer. Recently, this result has been generalized to cover inputs that are close to a coherent superposition of logarithmically many stabilizer states. The runtime of the classical simulation is governed by the stabilizer extent, which roughly measures how many stabilizer states are needed to approximate the state. An important open problem is to decide whether the extent is multiplicative under tensor products. An affirmative answer would yield an efficient algorithm for computing the extent of product inputs, while a negative result implies the existence of more efficient classical algorithms for simulating largescale quantum circuits. Here, we answer this question in the negative. Our result follows from very general properties of the set of stabilizer states, such as having a size that scales subexponentially in the dimension, and can thus be readily adapted to similar constructions for other resource theories.

Fredholm eigenvalues and quasiconformal geometry of polygons

An important open problem in geometric complex analysis is to establish algorithms for the explicit determination of the basic curvilinear and analytic functionals intrinsically connected with conformal and quasiconformal maps, such as their Teichmüller and Grunsky norms, Fredholm eigenvalues, and the quasireflection coefficient. This is important also for the potential theory but has not been solved even for convex polygons. This case has intrinsic interest in view of the connection of polygons with the geometry of the universal Teichmüller space and approximation theory. This survey extends our previous survey of 2005 and presents the newapproaches and recent essential progress in this field of geometric complex analysis and potential theory, having various important applications. Another new topic concerns quasireflections across finite collections of quasiintervals (to which the notion of Fredholm eigenvalues also can be extended).

Turing degrees in Polish spaces and decomposability of Borel functions

We give a partial answer to an important open problem in descriptive set theory, the Decomposability Conjecture for Borel functions on an analytic subset of a Polish space to a separable metrizable space. Our techniques employ deep results from effective descriptive set theory and recursion theory. In fact it is essential to extend several prominent results in recursion theory (e.g. the Shore–Slaman Join Theorem) to the setting of Polish spaces. As a by-product we give both positive and negative results on the Martin Conjecture on the degree preserving Borel functions between Polish spaces. Additionally we prove results about the transfinite version as well as the computable version of the Decomposability Conjecture.

Select to better learn: Fast and accurate deep learning using data selection from nonlinear manifolds

Finding a small subset of data whose linear combination spans other data points, also called column subset selection problem (CSSP), is an important open problem in computer science with many applications in computer vision and deep learning. There are some studies that solve CSSP in a polynomial time complexity w.r.t. the size of the original dataset. A simple and efficient selection algorithm with a linear complexity order, referred to as spectrum pursuit (SP), is proposed that pursuits spectral components of the dataset using available sample points. The proposed non-greedy algorithm aims to iteratively find K data samples whose span is close to that of the first K spectral components of entire data. SP has no parameter to be fine tuned and this desirable property makes it problem-independent. The simplicity of SP enables us to extend the underlying linear model to more complex models such as nonlinear manifolds and graph-based models. The nonlinear extension of SP is introduced as kernel-SP (KSP). The superiority of the proposed algorithms is demonstrated in a wide range of applications.

Select to better learn: Fast and accurate deep learning using data selection from nonlinear manifolds

Finding a small subset of data whose linear combination spans other data points, also called column subset selection problem (CSSP), is an important open problem in computer science with many applications in computer vision and deep learning. There are some studies that solve CSSP in a polynomial time complexity w.r.t. the size of the original dataset. A simple and efficient selection algorithm with a linear complexity order, referred to as spectrum pursuit (SP), is proposed that pursuits spectral components of the dataset using available sample points. The proposed non-greedy algorithm aims to iteratively find K data samples whose span is close to that of the first K spectral components of entire data. SP has no parameter to be fine tuned and this desirable property makes it problem-independent. The simplicity of SP enables us to extend the underlying linear model to more complex models such as nonlinear manifolds and graph-based models. The nonlinear extension of SP is introduced as kernel-SP (KSP). The superiority of the proposed algorithms is demonstrated in a wide range of applications.