Datasets:

nianlong

/

controllable-citation-generation

like 1

[ "We don't find particular differences in different classifiers, with the maximum being slightly weaker against this attack. These findings are in line Sharif et al.", "#OTHEREFR who report similar transferability for dodging attacks across different networks.", "Table 5 does not show any particularly evident trend in the transferability across architectures, showing that some transferability applies for each pair of models.", "Additionally, while it would seem that higher EER would lead to higher chance of attacks (it is easier to find accepted samples by chance), we find that baseline EER and success rate are only weakly correlated (r =-0.18).", "This suggests that the transferability properties rely on less intuitive combinations of both training datasets and architectures." ]

[]

[ "user template", "attacks" ]

{ "title": "Biometric Backdoors: A Poisoning Attack Against Unsupervised Template Updating", "abstract": "In this work, we investigate the concept of biometric backdoors: a template poisoning attack on biometric systems that allows adversaries to stealthily and effortlessly impersonate users in the long-term by exploiting the template update procedure. We show that such attacks can be carried out even by attackers with physical limitations (no digital access to the sensor) and zero knowledge of training data (they know neither decision boundaries nor user template). Based on the adversaries' own templates, they craft several intermediate samples that incrementally bridge the distance between their own template and the legitimate user's. As these adversarial samples are added to the template, the attacker is eventually accepted alongside the legitimate user. To avoid detection, we design the attack to minimize the number of rejected samples. We design our method to cope with the weak assumptions for the attacker and we evaluate the effectiveness of this approach on state-of-the-art face recognition pipelines based on deep neural networks. We find that in scenarios where the deep network is known, adversaries can successfully carry out the attack over 70% of cases with less than ten injection attempts. Even in black-box scenarios, we find that exploiting the transferability of adversarial samples from surrogate models can lead to successful attacks in around 15% of cases. Finally, we design a poisoning detection technique that leverages the consistent directionality of template updates in feature space to discriminate between legitimate and malicious updates. We evaluate such a countermeasure with a set of intra-user variability factors which may present the same directionality characteristics, obtaining equal error rates for the detection between 7-14% and leading to over 99% of attacks being detected after only two sample injections." }

{ "title": "Adversarial Generative Nets: Neural Network Attacks on State-of-the-Art Face Recognition.", "abstract": "In this paper we show that misclassification attacks against face-recognition systems based on deep neural networks (DNNs) are more dangerous than previously demonstrated, even in contexts where the adversary can manipulate only her physical appearance (versus directly manipulating the image input to the DNN). Specifically, we show how to create eyeglasses that, when worn, can succeed in targeted (impersonation) or untargeted (dodging) attacks while improving on previous work in one or more of three facets: (i) inconspicuousness to onlooking observers, which we test through a user study; (ii) robustness of the attack against proposed defenses; and (iii) scalability in the sense of decoupling eyeglass creation from the subject who will wear them, i.e., by creating \"universal\" sets of eyeglasses that facilitate misclassification. Central to these improvements are adversarial generative nets, a method we propose to generate physically realizable attack artifacts (here, eyeglasses) automatically." }

[ "These works do not target the creation of a covert timing authentication channel, but they focus on optimization problems for CAN traffic.", "Nonetheless, many recent research works have been focusing on using frame arrival time, i.e., the delays that we use to create a covert channel, in order to detect intrusions, e.g., #OTHEREFR and #OTHEREFR .", "By using Bloom filters #OTHEREFR , frame arrival time has been also combined with frame content to filter malicious activity in #OTHEREFR .", "More recently, frame periodicity has been exploited to extract clock skews which is used to create a unique fingerprint for each ECU due to physical imperfections in oscillators in #OTHEREFR . This sets room for physical fingerprinting of CAN nodes.", "The use of clock skews has been explored for fingerprinting computers for more than a decade by the work in #OTHEREFR and not surprisingly it was also applied to smart-phones #OTHEREFR ." ]

[ "All these works are exploiting the precision of the clock circuitry in the controller, which also stays at the core of our proposal here.", "To save bits from the data-field, other works have suggested the use of the identifier field, i.e., #OTHEREFR , #OTHEREFR , #OTHEREFR and #OTHEREFR , but this requires special care as the identifier field is critical for arbitration and also used for filtering purposes.", "An alternative to identify senders without compromising bits of the CAN frame is to use physical signal characteristics, e.g., #OTHEREFR , #OTHEREFR , #OTHEREFR , but these approaches may be vulnerable to small variations in bus impedance.", "Contribution in brief.", "In this work we create a covert authentication channel, which leaves the bits in the CAN frame unchanged, and increase its data rate by relying on efficient frame scheduling." ]

[ "time-triggered interrupts" ]

{ "title": "CANTO -- Covert AutheNtication with Timing channels over Optimized traffic flows for CAN", "abstract": "Abstract-Previous research works have endorsed the use of delays and clock skews for detecting intrusions or fingerprinting ECUs on the CAN bus. Similar techniques have been also proposed for establishing a time-covert cryptographic authentication channel, in this way cleverly removing the need for cryptographic material inside the limited payload of CAN frames. The main shortcoming of such works is the limited security level that can be achieved under normal CAN-bus traffic. In this work we endeavour to test the limits of the achievable security level by relying on optimization algorithms for scheduling CAN frames. Under practical bus allocations that are based on real-world scenarios, we are able to extract around 4-5 bits of authentication data from each frame which leads to an efficient intrusion detection and authentication mechanism. By accumulating covert channel data over several consecutive frames, we can achieve higher security levels that are in line with current security demands. To prove the correctness of our approach, we present experiments on state-of-the-art automotive-grade controllers (Infineon Aurix) and bus measurements with the use of industry standard tools, i.e., CANoe." }

{ "title": "Cloaking the Clock: Emulating Clock Skew in Controller Area Networks", "abstract": "Automobiles are equipped with Electronic Control Units (ECUs) that communicate via in-vehicle network protocol standards such as the Controller Area Network (CAN). These protocols were designed under the assumption that separating invehicle communications from external networks is sufficient for protection against cyber attacks. This assumption, however, has been shown to be invalid by recent attacks in which adversaries were able to infiltrate the in-vehicle network. Motivated by these attacks, intrusion detection systems (IDSs) have been proposed for in-vehicle networks that attempt to detect attacks by exploiting physical properties such as clock skew of an ECU. In this paper, we propose the cloaking attack, an intelligent masquerade attack in which an adversary modifies the timing of transmitted messages to match the clock skew of a targeted ECU. The attack leverages the fact that, while the clock skew is a physical property of each ECU that cannot be changed by the adversary, the estimation of the clock skew by other ECUs is based on the timing of network traffic, which, being a cyber component only, can be modified by an adversary. We implement the proposed cloaking attack and test it on two IDSs, namely, the current state-of-the-art IDS and its adaptation to the widelyused Network Time Protocol (NTP). We implement the cloaking attack on two hardware testbeds, a prototype and a real vehicle, and show that it is able to deceive both IDSs. We also introduce a new metric called the Maximum Slackness Index to quantify the effectiveness of a clock skew-based IDS in detecting masquerade attacks when the adversary is unable to precisely match the clock skew of the targeted ECU." }

[ "Observe as in Sect", "Here S ± (u) denote the support sets of a set S ⊆ R d :" ]

[ "The term involving Tr II is the same except for the terms ϕ", "If ρ is concentrated near 0, so that θ X (z) approximates the Dirac measure δ z (X), then D is small.", "Hence the formula is close to the corresponding formula in the black-and-white case." ]

[ "black-and-white case", "Theorem" ]

{ "title": "Estimation of Intrinsic Volumes from Digital Grey-Scale Images", "abstract": "Abstract Local algorithms are common tools for estimating intrinsic volumes from black-and-white digital images. However, these algorithms are typically biased in the design based setting, even when the resolution tends to infinity. Moreover, images recorded in practice are most often blurred grey-scale images rather than black-and-white. In this paper, an extended definition of local algorithms, applying directly to grey-scale images without thresholding, is suggested. We investigate the asymptotics of these new algorithms when the resolution tends to infinity and apply this to construct estimators for surface area and integrated mean curvature that are asymptotically unbiased in certain natural settings." }

{ "title": "Local digital algorithms for estimating the integrated mean curvature of r-regular sets", "abstract": "Abstract Consider the design based situation where an r-regular set is sampled on a random lattice. A fast algorithm for estimating the integrated mean curvature based on this observation is to use a weighted sum of 2 × · · ·× 2 configuration counts. We show that for a randomly translated lattice, no asymptotically unbiased estimator of this type exists in dimension greater than or equal to three, while for stationary isotropic lattices, asymptotically unbiased estimators are plenty. Both results follow from a general formula that we state and prove, describing the asymptotic behavior of hit-or-miss transforms of r-regular sets. Keywords Binary image · Design based set-up · Configurations · Mean curvature · r-regular sets · Hit-or-miss transform Mathematics Subject Classification (2010) 94A08 · 28A75 · 60D05" }

[ "This is satisfied by the weights in Table 5 .", "Proof.", "The assertion follows from a comparison of Corollary 5.2 with the Miles formulas (6).", "The weights in Table 5 fulfill the asserted condition since they fulfill (29) and Q 1 , . . .", ", Q 8 are the columns of the matrix Q." ]

[ "This follows since w (1) DQ 3 = 0 and w (9)] and hence that the estimator in the same theorem is unbiased.", "Thus, the weights in Table 5 are an optimal choice for isotropic Boolean models in R 3 with compact convex grains satisfying Condition 3.2.", "But in particular, they are also optimal based on the results of #OTHEREFR for the design based setting where an r-regular set is observed on a randomly translated and rotated lattice.", "6 Optimal algorithms for 3D Boolean models with balls as grains", "We now consider a stationary Boolean model whose grains are a.s." ]

[ "rotated lattice" ]

{ "title": "Local digital algorithms applied to Boolean models", "abstract": "We investigate the estimation of specific intrinsic volumes of stationary Boolean models by local digital algorithms; that is, by weighted sums of n × . . . × n configuration counts. We show that asymptotically unbiased estimators for the specific surface area or integrated mean curvature do not exist if the dimension is at least two or three, respectively. For 3-dimensional stationary, isotropic Boolean models, we derive asymptotically unbiased estimators for the specific surface area and integrated mean curvature. For a Boolean model with balls as grains we even obtain an asymptotically unbiased estimator for the specific Euler characteristic. This solves an open problem from [18] ." }

{ "title": "Local digital algorithms for estimating the integrated mean curvature of r-regular sets", "abstract": "Abstract Consider the design based situation where an r-regular set is sampled on a random lattice. A fast algorithm for estimating the integrated mean curvature based on this observation is to use a weighted sum of 2 × · · ·× 2 configuration counts. We show that for a randomly translated lattice, no asymptotically unbiased estimator of this type exists in dimension greater than or equal to three, while for stationary isotropic lattices, asymptotically unbiased estimators are plenty. Both results follow from a general formula that we state and prove, describing the asymptotic behavior of hit-or-miss transforms of r-regular sets. Keywords Binary image · Design based set-up · Configurations · Mean curvature · r-regular sets · Hit-or-miss transform Mathematics Subject Classification (2010) 94A08 · 28A75 · 60D05" }

[ "Several papers have explored how to represent divergence in big-step semantics.", "Leroy and Grall #OTHEREFR survey different approaches to representing divergence in coinductive big-step semantics, including divergence predicates, trace-based semantics, and taking the coinductive interpretation of standard big-step rules.", "They conclude that traditional divergence predicates are the most well-behaved, but increase the size of specifications by around 40%.", "The trace-based semantics of Leroy and Grall relies on concatenating infinite traces for accumulating the full trace of rules with multiple premises.", "Nakata and Uustalu #OTHEREFR propose a more elegant approach to accumulating traces based on 'peel' rules." ]

[ "In the partiality monad, functions either return a finitely delayed result or an infinite trace of delays.", "Piróg and Gibbons #OTHEREFR study the category theoretic foundations of the resumption monad.", "Related to the resumption monad is the interactive I/O monad by Hancock and Setzer #OTHEREFR for modeling the behaviour of possibly-diverging interactive programs.", "While it is possible to specify and reason about operational semantics by means of the partiality monad, it relies on the ability to express mixed recursive/corecursive functions in order to use it in proof assistants.", "Agda #OTHEREFR provides native support for such function definitions, while Coq does not." ]

[ "big-step semantics", "delay monad" ]

{ "title": "Flag-Based Big-Step Semantics", "abstract": "Structural operational semantic specifications come in different styles: small-step and big-step. A problem with the big-step style is that specifying divergence and abrupt termination gives rise to annoying duplication. We present a novel approach to representing divergence and abrupt termination in big-step semantics using status flags. This avoids the duplication problem, and uses fewer rules and premises for representing divergence than previous approaches in the literature." }

{ "title": "Normalization by Evaluation in the Delay Monad: A Case Study for Coinduction via Copatterns and Sized Types", "abstract": "In this paper, we present an Agda formalization of a normalizer for simply-typed lambda terms. The normalizer consists of two coinductively defined functions in the delay monad: One is a standard evaluator of lambda terms to closures, the other a type-directed reifier from values to η-long β -normal forms. Their composition, normalization-by-evaluation, is shown to be a total function a posteriori, using a standard logical-relations argument. The successful formalization serves as a proof-of-concept for coinductive programming and reasoning using sized types and copatterns, a new and presently experimental feature of Agda." }

[ "3.1 Find a Non-trivial Factor of x r − β", "We extend the square root algorithm in #OTHEREFR to show a deterministic algorithm, Algorithm 3.3, for finding a non-trivial factor of x r − β.", "Unlike other algorithms, such as the generalized Shanks's algorithm, Algorithm 3.3 does not require any r-th nonresidue as an input and the associated proofs do not assume any unproven hypothesis.", "Similar to #OTHEREFR , Algorithm 3.3 requires finding primitive roots of unity.", "It is obvious that finding an N -th primitive root of unity is not harder than finding an N -th nonresidue because, given an N -th nonresidue, an N -th primitive root of unity can be easily computed." ]

[ "Denote a fixed primitive k-th root of unity in F q by ζ k .", "(i) ζ 2 or ζ 3 when p ≡ 1 (mod 12).", "(ii) ζ 2·3 n +1 for n ≥ 1 when 2 · 3 n + 1 is a prime and p ≡ 13, 25 (mod 36).", "(iii) ζ r when q = r e t + 1 with t small.", "The arithmetic of the square root algorithm in #OTHEREFR is carried out over a specially constructed group, G α , which is isomorphic to F × q and a degenerated elliptic curve." ]

[ "primitive roots" ]

{ "title": "On Taking r-th Roots without r-th Nonresidues over Finite Fields and Its Applications", "abstract": "We first show a deterministic algorithm for taking r-th roots over F q without being given any r-th nonresidue, where F q is a finite field with q elements and r is a small prime such that r 2 divides of q−1. As applications, we illustrate deterministic algorithms over F q for constructing r-th nonresidues, constructing primitive elements, solving polynomial equations and computing elliptic curve \"n-th roots\", and a deterministic primality test for the generalized Proth numbers. All algorithms are proved without assuming any unproven hypothesis. They are efficient only if all the factors of q − 1 are small and some primitive roots of unity can be constructed efficiently over F q . In some cases, they are the fastest among the known deterministic algorithms." }

{ "title": "On Taking Square Roots without Quadratic Nonresidues over Finite Fields", "abstract": "Abstract. We present a novel idea to compute square roots over finite fields, without being given any quadratic nonresidue, and without assuming any unproven hypothesis. The algorithm is deterministic and the proof is elementary. In some cases, the square root algorithm runs inÕ(log 2 q) bit operations over finite fields with q elements. As an application, we construct a deterministic primality proving algorithm, which runs inÕ(log 3 N ) for some integers N ." }

[ "Below are some known cases that primitive roots of unity can be computed efficiently; see #OTHEREFR for more details. Let p be the characteristic of F q .", "Denote a fixed primitive k-th root of unity in F q by ζ k .", "(i) ζ 2 or ζ 3 when p ≡ 1 (mod 12).", "(ii) ζ 2·3 n +1 for n ≥ 1 when 2 · 3 n + 1 is a prime and p ≡ 13, 25 (mod 36).", "(iii) ζ r when q = r e t + 1 with t small." ]

[ "Taking square root is obviously equivalent to finding a non-trivial factor of x 2 − β.", "It is possible to formulate the algorithm in #OTHEREFR so that the arithmetic is carried out over the ring F q [x]/(x 2 − β) for factoring the polynomial x 2 − β.", "We generalize this idea and work on the ring F q [x]/(x r − β) in Algorithm 3.3.", "When r = 2, Algorithm 3.3 and the algorithm in #OTHEREFR are essentially the same.", "The \"problem\" of working on the ring F q [x]/(x r − β) is that there are zero divisors." ]

[ "square root algorithm" ]

{ "title": "On Taking r-th Roots without r-th Nonresidues over Finite Fields and Its Applications", "abstract": "We first show a deterministic algorithm for taking r-th roots over F q without being given any r-th nonresidue, where F q is a finite field with q elements and r is a small prime such that r 2 divides of q−1. As applications, we illustrate deterministic algorithms over F q for constructing r-th nonresidues, constructing primitive elements, solving polynomial equations and computing elliptic curve \"n-th roots\", and a deterministic primality test for the generalized Proth numbers. All algorithms are proved without assuming any unproven hypothesis. They are efficient only if all the factors of q − 1 are small and some primitive roots of unity can be constructed efficiently over F q . In some cases, they are the fastest among the known deterministic algorithms." }

{ "title": "On Taking Square Roots without Quadratic Nonresidues over Finite Fields", "abstract": "Abstract. We present a novel idea to compute square roots over finite fields, without being given any quadratic nonresidue, and without assuming any unproven hypothesis. The algorithm is deterministic and the proof is elementary. In some cases, the square root algorithm runs inÕ(log 2 q) bit operations over finite fields with q elements. As an application, we construct a deterministic primality proving algorithm, which runs inÕ(log 3 N ) for some integers N ." }

[ "(i) ζ 2 or ζ 3 when p ≡ 1 (mod 12).", "(ii) ζ 2·3 n +1 for n ≥ 1 when 2 · 3 n + 1 is a prime and p ≡ 13, 25 (mod 36).", "(iii) ζ r when q = r e t + 1 with t small.", "The arithmetic of the square root algorithm in #OTHEREFR is carried out over a specially constructed group, G α , which is isomorphic to F × q and a degenerated elliptic curve.", "Taking square root is obviously equivalent to finding a non-trivial factor of x 2 − β." ]

[ "We generalize this idea and work on the ring F q [x]/(x r − β) in Algorithm 3.3.", "When r = 2, Algorithm 3.3 and the algorithm in #OTHEREFR are essentially the same.", "The \"problem\" of working on the ring F q [x]/(x r − β) is that there are zero divisors.", "However, if we have a zero divisor f (x), then", "is a non-trivial factor of x r − β." ]

[ "arithmetic" ]

{ "title": "On Taking r-th Roots without r-th Nonresidues over Finite Fields and Its Applications", "abstract": "We first show a deterministic algorithm for taking r-th roots over F q without being given any r-th nonresidue, where F q is a finite field with q elements and r is a small prime such that r 2 divides of q−1. As applications, we illustrate deterministic algorithms over F q for constructing r-th nonresidues, constructing primitive elements, solving polynomial equations and computing elliptic curve \"n-th roots\", and a deterministic primality test for the generalized Proth numbers. All algorithms are proved without assuming any unproven hypothesis. They are efficient only if all the factors of q − 1 are small and some primitive roots of unity can be constructed efficiently over F q . In some cases, they are the fastest among the known deterministic algorithms." }

{ "title": "On Taking Square Roots without Quadratic Nonresidues over Finite Fields", "abstract": "Abstract. We present a novel idea to compute square roots over finite fields, without being given any quadratic nonresidue, and without assuming any unproven hypothesis. The algorithm is deterministic and the proof is elementary. In some cases, the square root algorithm runs inÕ(log 2 q) bit operations over finite fields with q elements. As an application, we construct a deterministic primality proving algorithm, which runs inÕ(log 3 N ) for some integers N ." }

[ "The generator network g θ (z, ε) takes a point z from latent space, and outputs a discrete-valued and symmetric graph adjacency tensor A and a discrete-valued node feature matrix X." ]

[ "We extend A and X to explicitly model the absence of edges and nodes by introducing a separate ghost-edge type and ghost-node type.", "This will enable us to encourage the generator to produce chemically valid molecular graphs as described below.", "Thus, we definẽ A ∈ {0, 1} n×n×(m+1) andX ∈ {0, 1} n×(d+1) .", "Each vectorà ij• , representing generated edges between nodes i and j, needs to be a member of the simplex", ", because only none or a single edge between i and j is allowed." ]

[ "MLP architecture" ]

{ "title": "Likelihood-Free Inference and Generation of Molecular Graphs", "abstract": "Recent methods for generating novel molecules use graph representations of molecules and employ various forms of graph convolutional neural networks for inference. However, training requires solving an expensive graph isomorphism problem, which previous approaches do not address or solve only approximately. In this work, we propose LF-MolGAN, a likelihood-free approach for de novo molecule generation that avoids explicitly computing a reconstruction loss. Our approach extends generative adversarial networks by including an adversarial cycle-consistency loss to implicitly enforce the reconstruction property. To capture properties unique to molecules, such as valence, we extend Graph Isomorphism Network to multi-graphs. To quantify the performance of models, we propose to compute the distance between distributions of physicochemical properties with the 1-Wasserstein distance. We demonstrate that LF-MolGAN more accurately learns the distribution over the space of molecules than all baselines. Moreover, it can be utilized for drug discovery by efficiently searching the space of molecules using molecules' continuous latent representation." }

{ "title": "MolGAN: An implicit generative model for small molecular graphs", "abstract": "Deep generative models for graph-structured data offer a new angle on the problem of chemical synthesis: by optimizing differentiable models that directly generate molecular graphs, it is possible to side-step expensive search procedures in the discrete and vast space of chemical structures. We introduce MolGAN, an implicit, likelihoodfree generative model for small molecular graphs that circumvents the need for expensive graph matching procedures or node ordering heuristics of previous likelihood-based methods. Our method adapts generative adversarial networks (GANs) to operate directly on graph-structured data. We combine our approach with a reinforcement learning objective to encourage the generation of molecules with specific desired chemical properties. In experiments on the QM9 chemical database, we demonstrate that our model is capable of generating close to 100% valid compounds. MolGAN compares favorably both to recent proposals that use string-based (SMILES) representations of molecules and to a likelihoodbased method that directly generates graphs, albeit being susceptible to mode collapse." }

[ "In contrast to traditional approaches, which first try to understand the question, and then evaluate the query, in gAnswer the intention of the query is modeled in a structured way, which leads to a subgraph matching problem.", "Secondly, QAKiS #OTHEREFR is QA system over structured knowledge bases such as DBpedia that makes use of relational patterns which capture different ways to express a certain relation in a natural language in order to construct a target-language (SPARQL) query.", "Further, Platypus #OTHEREFR ) is a QA system on Wikidata.", "It represents questions in an internal format related to dependency-based compositional semantics which allows for question decomposition and language independence.", "The platform can answer complex questions in several languages by using hybrid grammatical and template-based techniques." ]

[ "WDAqua puts more importance on word semantics than on the syntax of the user query, and follows a processes of query expansion, SPARQL construction, query ranking and then making an answer decision.", "For the evaluation of QA systems, several benchmarks have been proposed such as WebQuestions #OTHEREFR or SimpleQuestions #OTHEREFR .", "However, the most popular benchmarks in the Semantic Web field arise from the QALD evaluation campaign #OTHEREFR .", "The recent QALD7 evaluation campaign includes task 4: \"English question answering over Wikidata\" 2 which serves as basis to compile our evaluation dataset." ]

[ "underlying knowledge base" ]

{ "title": "A Comparative Evaluation of Visual and Natural Language Question Answering Over Linked Data", "abstract": "With the growing number and size of Linked Data datasets, it is crucial to make the data accessible and useful for users without knowledge of formal query languages. Two approaches towards this goal are knowledge graph visualization and natural language interfaces. Here, we investigate specifically question answering (QA) over Linked Data by comparing a diagrammatic visual approach with existing natural language-based systems. Given a QA benchmark (QALD7), we evaluate a visual method which is based on iteratively creating diagrams until the answer is found, against four QA systems that have natural language queries as input. Besides other benefits, the visual approach provides higher performance, but also requires more manual input. The results indicate that the methods can be used complementary, and that such a combination has a large positive impact on QA performance, and also facilitates additional features such as data exploration." }

{ "title": "Towards a Question Answering System over the Semantic Web", "abstract": "Abstract. Thanks to the development of the Semantic Web, a lot of new structured data has become available on the Web in the form of knowledge bases (KBs). Making this valuable data accessible and usable for end-users is one of the main goals of Question Answering (QA) over KBs. Most current QA systems query one KB, in one language (namely English). The existing approaches are not designed to be easily adaptable to new KBs and languages. We first introduce a new approach for translating natural language questions to SPARQL queries. It is able to query several KBs simultaneously, in different languages, and can easily be ported to other KBs and languages. In our evaluation, the impact of our approach is proven using 5 different well-known and large KBs: Wikidata, DBpedia, MusicBrainz, DBLP and Freebase as well as 5 different languages namely English, German, French, Italian and Spanish. Second, we show how we integrated our approach, to make it easily accessible by the research community and by end-users. To summarize, we provided a conceptional solution for multilingual, KB-agnostic Question Answering over the Semantic Web. The provided first approximation validates this concept." }

[ "In #OTHEREFR the safety margin to the decision thresholds of the received Phase-Shift Keying (PSK) symbols is maximized subject to a relaxed 1-bit constraint using linear programming for flatfading channels and extended in #OTHEREFR for frequency-selective channels.", "The same optimization problem was solved by the Branch-and Bound algorithm in #OTHEREFR for the special case of 4-PSK.", "To the best of our knowledge, the only works that have considered the case of coarsely QCE transmit signals are #OTHEREFR - #OTHEREFR .", "In #OTHEREFR , we propose a symbol-wise MSE precoder based on gradient-descent under a strict CE constraint or a relaxed polygon constraint.", "In #OTHEREFR , the authors extend the method in #OTHEREFR to fit the context of QCE transmit signals." ]

[ "The contribution in #OTHEREFR addresses the task of QCE precoding in the context of using a single common PA and separate digital phase shifters for the antenna front-ends.", "The optimization problem consists of designing the QCE precoder while minimizing the MUI, and the idea of constructive interference, #OTHEREFR , #OTHEREFR , is not exploited as in our work.", "The concept of QCE precoding and general constellations for flatfading channels is studied in this contribution and is extended partially to frequency-selective channels in #OTHEREFR and #OTHEREFR .", "It is worth mentioning that the QCE precoding can be combined with appropriate pulse shaping strategies as in #OTHEREFR and #OTHEREFR to ensure an efficient spectral confinement.", "In #OTHEREFR , it was shown that CE precoding is still power efficient even when considering time-based processing." ]

[ "precoder design" ]

{ "title": "Quantized Constant Envelope Precoding With PSK and QAM Signaling", "abstract": "Coarsely quantized massive multiple-input multiple-output (MIMO) systems are gaining more interest due to their power efficiency. We present a new precoding technique to mitigate the multi-user interference and the quantization distortions in a downlink multi-user MIMO system with coarsely quantized constant envelope (QCE) signals at the transmitter. The transmit signal vector is optimized for every desired received vector taking into account a relaxed version of the QCE constraint. The optimization is based on maximizing the safety margin to the decision thresholds of the receiver constellation modulation. Due to the linear property of the objective function and the constraints, the optimization problem is formulated as a linear programming problem. The simulation results show a significant gain in terms of the uncoded bit error rate compared to the existing precoding techniques. Index Terms-Constant envelope, coarse quantization, constructive interference, downlink massive multi-user MIMO, precoding." }

{ "title": "Quantized Precoding for Multi-Antenna Downlink Channels with MAGIQ", "abstract": "A multi-antenna, greedy, iterative, and quantized (MAGIQ) precoding algorithm is proposed for downlink channels. MAGIQ allows a straightforward integration with orthogonal frequency-division multiplexing (OFDM). MAGIQ is compared to three existing algorithms in terms of information rates and complexity: quantized linear precoding (QLP), SQUID, and an ADMM-based algorithm. The information rate is measured by using a lower bound for finite modulation sets, and the complexity is measured by the number of multiplications and comparisons. MAGIQ and ADMM achieve similar information rates with similar complexity for Rayleigh flat-fading channels and one-bit quantization per real dimension, and they outperform QLP and SQUID for higher order modulation." }

[ "Hence, to prove Theorem 1, we need to show that the entry on the mth row and on the nth column of the RHS of (62) equals the RHS of (42) if we use (53) to evaluate C e .", "Define u a = a/Δ and v b = b/Δ.", "Then, we can write the entry on the mth row and nth column of the RHS of (62) as follows:", "Here, to obtain (63), we replaced [C e ] m,n with #OTHEREFR .", "To obtain (64), we used that (63) is a two-dimensional Riemann sum, which, by definition, can be written as the twodimensional integral in (64)." ]

[]

[ "proof" ]

{ "title": "Linear Precoding With Low-Resolution DACs for Massive MU-MIMO-OFDM Downlink", "abstract": "We consider the downlink of a massive multiuser (MU) multiple-input multiple-output (MIMO) system in which the base station (BS) is equipped with low-resolution digital-to-analog converters (DACs). In contrast to most existing results, we assume that the system operates over a frequencyselective wideband channel and uses orthogonal frequency division multiplexing (OFDM) to simplify equalization at the user equipments (UEs). Furthermore, we consider the practically relevant case of oversampling DACs. We theoretically analyze the uncoded bit error rate (BER) performance with linear precoders (e.g., zero forcing) and quadrature phase-shift keying using Bussgang's theorem. We also develop a lower bound on the information-theoretic sum-rate throughput achievable with Gaussian inputs, which can be evaluated in closed form for the case of 1-bit DACs. For the case of multi-bit DACs, we derive approximate, yet accurate, expressions for the distortion caused by low-precision DACs, which can be used to establish the lower bounds on the corresponding sum-rate throughput. Our results demonstrate that, for a massive MU-MIMO-OFDM system with a 128-antenna BS serving 16 UEs, only 3-4 DAC bits are required to achieve an uncoded BER of 10 −4 with a negligible performance loss compared to the infinite-resolution case at the cost of additional out-of-band emissions. Furthermore, our results highlight the importance of considering the inherent spatial and temporal correlations caused by low-precision DACs. Index Terms-Massive multiuser multiple-input multiple output, low-resolution digital-to-analog converter, linear precoding, orthogonal frequency-division multiplexing, Bussgang's theorem." }

{ "title": "Quantized Precoding for Multi-Antenna Downlink Channels with MAGIQ", "abstract": "A multi-antenna, greedy, iterative, and quantized (MAGIQ) precoding algorithm is proposed for downlink channels. MAGIQ allows a straightforward integration with orthogonal frequency-division multiplexing (OFDM). MAGIQ is compared to three existing algorithms in terms of information rates and complexity: quantized linear precoding (QLP), SQUID, and an ADMM-based algorithm. The information rate is measured by using a lower bound for finite modulation sets, and the complexity is measured by the number of multiplications and comparisons. MAGIQ and ADMM achieve similar information rates with similar complexity for Rayleigh flat-fading channels and one-bit quantization per real dimension, and they outperform QLP and SQUID for higher order modulation." }

[ "The author proves that it is possible to robustly learn individual hypotheses for each source, but a single shared hypothesis cannot be learned robustly.", "For the specific case that all data distributions are identical, the setup matches ours, though only for binary classification in the realizable case, and with a different adversarial model.", "In a similar setting, Mahloujifar et al.", "(2019) show, in particular, that an adversary can increase the probability of any \"bad property\" of the learned hypothesis by a term at least proportional to the fraction of manipulated sources.", "These results differ from ours, by their assumption that different sources have different distributions, which renders the learning problem much harder." ]

[ "The main difference to our work is that their proposed method crucially relies on a trusted subset of the data being known to the learner.", "Their adversary is also weaker, as it cannot influence the data points directly, but only change the distribution from which they are sampled, and the work also does not provide finite sample guarantees.", "There are a number of classic results on the fundamental limits of PAC learning from a single labelled set of samples, a fraction of which can be arbitrarily corrupted, e.g. #OTHEREFR .", "We compare our results against this classic scenario in Section 4.1.", "Another related general direction is the research on Byzantine-resilient distributed learning, which has seen significant interest recently, e.g." ]

[ "datasets" ]

{ "title": "On the Sample Complexity of Adversarial Multi-Source PAC Learning", "abstract": "We study the problem of learning from multiple untrusted data sources, a scenario of increasing practical relevance given the recent emergence of crowdsourcing and collaborative learning paradigms. Specifically, we analyze the situation in which a learning system obtains datasets from multiple sources, some of which might be biased or even adversarially perturbed. It is known that in the single-source case, an adversary with the power to corrupt a fixed fraction of the training data can prevent PAC-learnability, that is, even in the limit of infinitely much training data, no learning system can approach the optimal test error. In this work we show that, surprisingly, the same is not true in the multi-source setting, where the adversary can arbitrarily corrupt a fixed fraction of the data sources. Our main results are a generalization bound that provides finite-sample guarantees for this learning setting, as well as corresponding lower bounds. Besides establishing PAC-learnability our results also show that in a cooperative learning setting sharing data with other parties has provable benefits, even if some participants are malicious." }

{ "title": "Robust Learning from Untrusted Sources", "abstract": "Modern machine learning methods often require more data for training than a single expert can provide. Therefore, it has become a standard procedure to collect data from external sources, e.g. via crowdsourcing. Unfortunately, the quality of these sources is not always guaranteed. As additional complications, the data might be stored in a distributed way, or might even have to remain private. In this work, we address the question of how to learn robustly in such scenarios. Studying the problem through the lens of statistical learning theory, we derive a procedure that allows for learning from all available sources, yet automatically suppresses irrelevant or corrupted data. We show by extensive experiments that our method provides significant improvements over alternative approaches from robust statistics and distributed optimization." }

[ "Before introducing the class of localized SLCs, we present a simple example for which the QI framework fails to capture an \"obvious\" controller with localized structure, but for which the SLA can.", "This example also serves to illustrate the importance of locality in achieving scalability of controller implementation." ]

[ "Example 1: Consider the optimal control problem as follows: #OTHEREFR with disturbance w [t] i.i. d ∼ N(0, I) .", "We assume full state feedback, i.e., the control action at time t can be expressed as", "for some function f .", "An optimal control policy u for this linear quadratic regulator (LQR) problem is easily seen to be given by u [t] = −Ax [t] .", "Further, suppose that the state matrix A is sparse and let its support define the adjacency matrix of a graph G for which we identify the ith node with the corresponding state/control pair (x i , u i )." ]

[ "controller synthesis" ]

{ "title": "A System-Level Approach to Controller Synthesis", "abstract": "Abstract-Biological and advanced cyber-physical control systems often have limited, sparse, uncertain, and distributed communication and computing in addition to sensing and actuation. Fortunately, the corresponding plants and performance requirements are also sparse and structured, and this must be exploited to make constrained controller design feasible and tractable. We introduce a new \"system level\" (SL) approach involving three complementary SL elements. SL parameterizations (SLPs) provide an alternative to the Youla parameterization of all stabilizing controllers and the responses they achieve, and combine with SL constraints (SLCs) to parameterize the largest known class of constrained stabilizing controllers that admit a convex characterization, generalizing quadratic invariance. SLPs also lead to a generalization of detectability and stabilizability, suggesting the existence of a rich separation structure, that when combined with SLCs is naturally applicable to structurally constrained controllers and systems. We further provide a catalog of useful SLCs, most importantly including sparsity, delay, and locality constraints on both communication and computing internal to the controller, and external system performance. Finally, we formulate SL synthesis problems, which define the broadest known class of constrained optimal control problems that can be solved using convex programming. Index Terms-Constrained and structured optimal control, decentralized control, large-scale systems." }

{ "title": "Separable and Localized System-Level Synthesis for Large-Scale Systems", "abstract": "Abstract-A major challenge faced in the design of largescale cyber-physical systems, such as power systems, the Internet of Things or intelligent transportation systems, is that traditional distributed optimal control methods do not scale gracefully, neither in controller synthesis nor in controller implementation, to systems composed of a large number (e.g., on the order of billions) of interacting subsystems. This paper shows that this challenge can now be addressed by leveraging the recently introduced systemlevel approach (SLA) to controller synthesis. In particular, in the context of the SLA, we define suitable notions of separability for control objective functions and system constraints such that the global optimization problem (or iterate update problems of a distributed optimization algorithm) can be decomposed into parallel subproblems. We then further show that if additional locality (i.e., sparsity) constraints are imposed, then these subproblems can be solved using local models and local decision variables. The SLA is essential to maintain the convexity of the aforementioned problems under locality constraints. As a consequence, the resulting synthesis methods have O(1) complexity relative to the size of the global system. We further show that many optimal control problems of interest, such as (localized) LQR and LQG, H 2 optimal control with joint actuator and sensor regularization, and (localized) mixed H 2 /L 1 optimal control problems, satisfy these notions of separability, and use these problems to explore tradeoffs in performance, actuator, and sensing density, and average versus worst-case performance for a large-scale power inspired system. Index Terms-Constrained and structured optimal control, decentralized control, large-scale systems, systemlevel synthesis (SLS)." }

[ "In this paper, we defined and analyzed the SLA to the controller synthesis, which consists of three elements: SLPs, SLCs, and SLS problems.", "We showed that all achievable and stable system responses can be characterized via the SLPs given in Theorems 1 and 2.", "We further showed that these system responses could be used to parameterize internally stabilizing controllers that achieved them, and proposed a novel controller implementation #OTHEREFR .", "We then argued that this novel controller implementation had the important benefit of allowing for SLCs to be naturally imposed on it, and showed in Section IV that using this controller structure and SLCs, we can characterize the broadest known class of constrained internally stabilizing controllers that admit a convex representation.", "Finally, we combined SLPs and SLCs to formulate the SLS problem, and showed that it recovered as a special case many well-studied constrained optimal controller synthesis problems from the literature." ]

[]

[ "controller synthesis", "largescale networked systems" ]

{ "title": "A System-Level Approach to Controller Synthesis", "abstract": "Abstract-Biological and advanced cyber-physical control systems often have limited, sparse, uncertain, and distributed communication and computing in addition to sensing and actuation. Fortunately, the corresponding plants and performance requirements are also sparse and structured, and this must be exploited to make constrained controller design feasible and tractable. We introduce a new \"system level\" (SL) approach involving three complementary SL elements. SL parameterizations (SLPs) provide an alternative to the Youla parameterization of all stabilizing controllers and the responses they achieve, and combine with SL constraints (SLCs) to parameterize the largest known class of constrained stabilizing controllers that admit a convex characterization, generalizing quadratic invariance. SLPs also lead to a generalization of detectability and stabilizability, suggesting the existence of a rich separation structure, that when combined with SLCs is naturally applicable to structurally constrained controllers and systems. We further provide a catalog of useful SLCs, most importantly including sparsity, delay, and locality constraints on both communication and computing internal to the controller, and external system performance. Finally, we formulate SL synthesis problems, which define the broadest known class of constrained optimal control problems that can be solved using convex programming. Index Terms-Constrained and structured optimal control, decentralized control, large-scale systems." }

{ "title": "Separable and Localized System-Level Synthesis for Large-Scale Systems", "abstract": "Abstract-A major challenge faced in the design of largescale cyber-physical systems, such as power systems, the Internet of Things or intelligent transportation systems, is that traditional distributed optimal control methods do not scale gracefully, neither in controller synthesis nor in controller implementation, to systems composed of a large number (e.g., on the order of billions) of interacting subsystems. This paper shows that this challenge can now be addressed by leveraging the recently introduced systemlevel approach (SLA) to controller synthesis. In particular, in the context of the SLA, we define suitable notions of separability for control objective functions and system constraints such that the global optimization problem (or iterate update problems of a distributed optimization algorithm) can be decomposed into parallel subproblems. We then further show that if additional locality (i.e., sparsity) constraints are imposed, then these subproblems can be solved using local models and local decision variables. The SLA is essential to maintain the convexity of the aforementioned problems under locality constraints. As a consequence, the resulting synthesis methods have O(1) complexity relative to the size of the global system. We further show that many optimal control problems of interest, such as (localized) LQR and LQG, H 2 optimal control with joint actuator and sensor regularization, and (localized) mixed H 2 /L 1 optimal control problems, satisfy these notions of separability, and use these problems to explore tradeoffs in performance, actuator, and sensing density, and average versus worst-case performance for a large-scale power inspired system. Index Terms-Constrained and structured optimal control, decentralized control, large-scale systems, systemlevel synthesis (SLS)." }

[ "In this paper, we generalized the SLS parameterization of LTI ∞ -stabilizing controllers, as well as a robust counterpart, to systems described by bounded and causal linear operators.", "This extension, along with a simple algebraic transformation, allowed us to leverage tools from L 1 robust control to derive necessary and sufficient conditions for the robust performance of an uncertain system in terms of convex constraints on the system response.", "We argued that these conditions remain necessary and sufficient when additional structure, such as that induced by delay, sparsity, and locality constraints, are imposed on the system response, so long as the uncertain elements (∆ A , ∆ B ) remained unstructured." ]

[ "More importantly, we believe that the results in this paper open up a wide and exciting range of future research directions, including but not limited to, deriving analogous results for H ∞ optimal control, revisiting the structured singular value, µ-synthesis, and structured small gain theorems from a system level perspective, and perhaps most exciting, exploring the interplay between closed loop locality constraints and additional structure in the dynamic uncertainty (∆ A , ∆ B ).", "Further, it is of interest to see if these tighter conditions can be used to derive interpretable bounds on the degradation in performance of a robust controller as a function of the norm bound ε on the uncertainty [∆ A , ∆ B ] ∞→ ∞ , as coarser and more conservative versions of such bounds have proved crucial in combining machine learning and robust control techniques #OTHEREFR ." ]

[ "resulting robust performance", "distributed synthesis techniques" ]

{ "title": "Robust Performance Guarantees for System Level Synthesis", "abstract": "We generalize the System Level Synthesis (SLS) framework to systems defined by bounded causal linear operators, and use this parameterization to make connections between robust SLS and classical results from the robust control literature. In particular, by leveraging results from L 1 robust control, we show that necessary and sufficient conditions for robust performance with respect to causal bounded linear uncertainty in the system dynamics can be translated into convex constraints on the system responses. We exploit this connection to show that these conditions naturally allow for the incorporation of delay, sparsity, and locality constraints on the system responses and resulting controller implementation, allowing these methods to be applied to large-scale distributed control problems -to the best of our knowledge, these are the first such robust performance guarantees for distributed control systems." }

{ "title": "Separable and Localized System-Level Synthesis for Large-Scale Systems", "abstract": "Abstract-A major challenge faced in the design of largescale cyber-physical systems, such as power systems, the Internet of Things or intelligent transportation systems, is that traditional distributed optimal control methods do not scale gracefully, neither in controller synthesis nor in controller implementation, to systems composed of a large number (e.g., on the order of billions) of interacting subsystems. This paper shows that this challenge can now be addressed by leveraging the recently introduced systemlevel approach (SLA) to controller synthesis. In particular, in the context of the SLA, we define suitable notions of separability for control objective functions and system constraints such that the global optimization problem (or iterate update problems of a distributed optimization algorithm) can be decomposed into parallel subproblems. We then further show that if additional locality (i.e., sparsity) constraints are imposed, then these subproblems can be solved using local models and local decision variables. The SLA is essential to maintain the convexity of the aforementioned problems under locality constraints. As a consequence, the resulting synthesis methods have O(1) complexity relative to the size of the global system. We further show that many optimal control problems of interest, such as (localized) LQR and LQG, H 2 optimal control with joint actuator and sensor regularization, and (localized) mixed H 2 /L 1 optimal control problems, satisfy these notions of separability, and use these problems to explore tradeoffs in performance, actuator, and sensing density, and average versus worst-case performance for a large-scale power inspired system. Index Terms-Constrained and structured optimal control, decentralized control, large-scale systems, systemlevel synthesis (SLS)." }