Search

Symmetries of Seiberg-Witten (SW) geometries capture intricate physical aspects of the underlying 4d $\mathcal{N} = 2$ field theories. For rank-one theories, these geometries are rational elliptic surfaces whose automorphism group is a semi-direct product between the Coulomb branch (CB) symmetries and the Mordell-Weil group. We study quotients of the SW geometry by subgroups of its automorphism group, which most naturally become gaugings of discrete 0- and 1-form symmetries. Yet, new interpretations of these surgeries become evident when considering 5d $\mathcal{N}=1$ superconformal field theories. There, certain CB symmetries are related to symmetries of the corresponding $(p,q)$-brane web and, as a result, CB surgeries can give rise to (fractional) S-folds. Another novel interpretation of these quotients is the folding across dimensions: circle compactifications of the 5d $E_{2N_f + 1}$ Seiberg theories lead in the infrared to two copies of locally indistinguishable 4d SU(2) SQCD theories with $N_f$ fundamental flavours. This extends earlier results on holonomy saddles, while also reproducing detailed computations of 5d BPS spectra and predicting new 5d and 6d BPS quivers. Finally, we argue that the semi-direct product structure of the automorphism group of the SW geometry includes mixed 't Hooft anomalies between the 0- and 1-form symmetries, and we also present some new results on non-cyclic CB symmetries., Comment: 52 pages + appendices, 10 figures. v2: minor changes, references added

The widespread adoption of large-scale machine learning models in recent years highlights the need for distributed computing for efficiency and scalability. This work introduces a novel distributed machine learning paradigm -- \emph{consensus learning} -- which combines classical ensemble methods with consensus protocols deployed in peer-to-peer systems. These algorithms consist of two phases: first, participants develop their models and submit predictions for any new data inputs; second, the individual predictions are used as inputs for a communication phase, which is governed by a consensus protocol. Consensus learning ensures user data privacy, while also inheriting the safety measures against Byzantine attacks from the underlying consensus mechanism. We provide a detailed theoretical analysis for a particular consensus protocol and compare the performance of the consensus learning ensemble with centralised ensemble learning algorithms. The discussion is supplemented by various numerical simulations, which describe the robustness of the algorithms against Byzantine participants., Comment: 27 pages plus appendices, 8 figures

We study how the global structure of rank-one 4d $\mathcal{N}=2$ supersymmetric field theories is encoded into global aspects of the Seiberg-Witten elliptic fibration. Starting with the prototypical example of the $\mathfrak{su}(2)$ gauge theory, we distinguish between relative and absolute Seiberg-Witten curves. For instance, we discuss in detail the three distinct absolute curves for the $SU(2)$ and $SO(3)_\pm$ 4d $\mathcal{N}=2$ gauge theories. We propose that the $1$-form symmetry of an absolute theory is isomorphic to a torsion subgroup of the Mordell-Weil group of sections of the absolute curve, while the full defect group of the theory is encoded in the torsion sections of a so-called relative curve. We explicitly show that the relative and absolute curves are related by isogenies (that is, homomorphisms of elliptic curves) generated by torsion sections -- hence, gauging a one-form symmetry corresponds to composing isogenies between Seiberg-Witten curves. We apply this approach to Kaluza-Klein (KK) 4d $\mathcal{N}=2$ theories that arise from toroidal compactifications of 5d and 6d SCFTs to four dimensions, uncovering an intricate pattern of 4d global structures obtained by gauging discrete $0$-form and/or $1$-form symmetries. Incidentally, we propose a 6d BPS quiver for the 6d M-string theory on $\mathbb{R}^4\times T^2$., Comment: 62 pages plus appendix; v2: added section 2.3 on non-invertible symmetry in the SO(3) theory, fixed typos and references; v3: minor rewriting of appendix A.5

We study 5d $\mathcal{N}=1$ supersymmetric field theories on closed five-manifolds $\mathcal{M}_5$ which are principal circle bundles over simply-connected K\"ahler four-manifolds, $\mathcal{M}_4$, equipped with the Donaldson-Witten twist. We propose a new approach to compute the supersymmetric partition function on $\mathcal{M}_5$ through the insertion of a fibering operator, which introduces a non-trivial fibration over $\mathcal{M}_4$, in the 4d topologically twisted field theory. We determine the so-called Coulomb branch partition function on any such $\mathcal{M}_5$, which is conjectured to be the holomorphic `integrand' of the full partition function. We precisely match the low-energy effective field theory approach to explicit one-loop computations, and we discuss the effect of non-perturbative 5d BPS particles in this context. When $\mathcal{M}_4$ is toric, we also reconstruct our Coulomb branch partition function by appropriately gluing Nekrasov partition functions. As a by-product of our analysis, we provide strong new evidence for the validity of the Lockhart-Vafa formula for the five-sphere partition function., Comment: 65 pages plus appendix; v2: corrected typos, JHEP version

We study the Coulomb branches of the rank-one 4d $\mathcal{N} = 2$ quantum field theories, including the KK theories obtained from the circle compactification of the 5d $\mathcal{N}= 1$ $E_n$ Seiberg theories. The focus is set on the relation between their Seiberg-Witten geometries and rational elliptic surfaces, with more attention being given to the modular surfaces, which we completely classify using the classification of subgroups of the modular group ${\rm PSL}(2,\mathbb{Z})$. We derive closed-form expressions for the modular functions for all congruence and some of the non-congruence subgroups associated with these geometries. Moreover, we show how the BPS quivers of these theories can be determined directly from the fundamental domains of the monodromy groups and study how changes of these domains can be interpreted as quiver mutations. This approach can also be generalized to theories whose Coulomb branches contain `undeformable' singularities, leading to known quivers of such theories., Comment: 63 pages plus appendix, many figures; v2: added references, corrected typos

Web automation holds the potential to revolutionize how users interact with the digital world, offering unparalleled assistance and simplifying tasks via sophisticated computational methods. Central to this evolution is the web element nomination task, which entails identifying unique elements on webpages. Unfortunately, the development of algorithmic designs for web automation is hampered by the scarcity of comprehensive and realistic datasets that reflect the complexity faced by real-world applications on the Web. To address this, we introduce the Klarna Product Page Dataset, a comprehensive and diverse collection of webpages that surpasses existing datasets in richness and variety. The dataset features 51,701 manually labeled product pages from 8,175 e-commerce websites across eight geographic regions, accompanied by a dataset of rendered page screenshots. To initiate research on the Klarna Product Page Dataset, we empirically benchmark a range of Graph Neural Networks (GNNs) on the web element nomination task. We make three important contributions. First, we found that a simple Convolutional GNN (GCN) outperforms complex state-of-the-art nomination methods. Second, we introduce a training refinement procedure that involves identifying a small number of relevant elements from each page using the aforementioned GCN. These elements are then passed to a large language model for the final nomination. This procedure significantly improves the nomination accuracy by 16.8 percentage points on our challenging dataset, without any need for fine-tuning. Finally, in response to another prevalent challenge in this field - the abundance of training methodologies suitable for element nomination - we introduce the Challenge Nomination Training Procedure, a novel training approach that further boosts nomination accuracy., Comment: 12 pages, 8 figures, 3 tables, under review

Search engines answer users' queries by listing relevant items (e.g. documents, songs, products, web pages, ...). These engines rely on algorithms that learn to rank items so as to present an ordered list maximizing the probability that it contains relevant item. The main challenge in the design of learning-to-rank algorithms stems from the fact that queries often have different meanings for different users. In absence of any contextual information about the query, one often has to adhere to the {\it diversity} principle, i.e., to return a list covering the various possible topics or meanings of the query. To formalize this learning-to-rank problem, we propose a natural model where (i) items are categorized into topics, (ii) users find items relevant only if they match the topic of their query, and (iii) the engine is not aware of the topic of an arriving query, nor of the frequency at which queries related to various topics arrive, nor of the topic-dependent click-through-rates of the items. For this problem, we devise LDR (Learning Diverse Rankings), an algorithm that efficiently learns the optimal list based on users' feedback only. We show that after $T$ queries, the regret of LDR scales as $O((N-L)\log(T))$ where $N$ is the number of all items. We further establish that this scaling cannot be improved, i.e., LDR is order optimal. Finally, using numerical experiments on both artificial and real-world data, we illustrate the superiority of LDR compared to existing learning-to-rank algorithms., Comment: 26 pages, 4 Figures, accepted in ACM SIGMETRICS 2018

The simplest non-trivial 5d superconformal field theories (SCFT) are the famous rank-one theories with $E_n$ flavour symmetry. We study their $U$-plane, which is the one-dimensional Coulomb branch of the theory on $\mathbb{R}^4 \times S^1$. The total space of the Seiberg-Witten (SW) geometry -- the $E_n$ SW curve fibered over the $U$-plane -- is described as a rational elliptic surface with a singular fiber of type $I_{9-n}$ at infinity. A classification of all possible Coulomb branch configurations, for the $E_n$ theories and their 4d descendants, is given by Persson's classification of rational elliptic surfaces. We show that the global form of the flavour symmetry group is encoded in the Mordell-Weil group of the SW elliptic fibration. We study in detail many special points in parameters space, such as points where the flavour symmetry enhances, and/or where Argyres-Douglas and Minahan-Nemeschansky theories appear. In a number of important instances, including in the massless limit, the $U$-plane is a modular curve, and we use modularity to investigate aspects of the low-energy physics, such as the spectrum of light particles at strong coupling and the associated BPS quivers. We also study the gravitational couplings on the $U$-plane, matching the infrared expectation for the couplings $A(U)$ and $B(U)$ to the UV computation using the Nekrasov partition function., Comment: 137 pages plus appendix, many figures. v2: added references, corrected typos and imprecisions; v3: corrected typos and references. SciPost version

This paper introduces and addresses a wide class of stochastic bandit problems where the function mapping the arm to the corresponding reward exhibits some known structural properties. Most existing structures (e.g. linear, Lipschitz, unimodal, combinatorial, dueling, ...) are covered by our framework. We derive an asymptotic instance-specific regret lower bound for these problems, and develop OSSB, an algorithm whose regret matches this fundamental limit. OSSB is not based on the classical principle of "optimism in the face of uncertainty" or on Thompson sampling, and rather aims at matching the minimal exploration rates of sub-optimal arms as characterized in the derivation of the regret lower bound. We illustrate the efficiency of OSSB using numerical experiments in the case of the linear bandit problem and show that OSSB outperforms existing algorithms, including Thompson sampling., Comment: 13 pages, NIPS 2017

We consider stochastic multi-armed bandit problems where the expected reward is a Lipschitz function of the arm, and where the set of arms is either discrete or continuous. For discrete Lipschitz bandits, we derive asymptotic problem specific lower bounds for the regret satisfied by any algorithm, and propose OSLB and CKL-UCB, two algorithms that efficiently exploit the Lipschitz structure of the problem. In fact, we prove that OSLB is asymptotically optimal, as its asymptotic regret matches the lower bound. The regret analysis of our algorithms relies on a new concentration inequality for weighted sums of KL divergences between the empirical distributions of rewards and their true distributions. For continuous Lipschitz bandits, we propose to first discretize the action space, and then apply OSLB or CKL-UCB, algorithms that provably exploit the structure efficiently. This approach is shown, through numerical experiments, to significantly outperform existing algorithms that directly deal with the continuous set of arms. Finally the results and algorithms are extended to contextual bandits with similarities., Comment: COLT 2014