Search

Transformers have revolutionized machine learning with their simple yet effective architecture. Pre-training Transformers on massive text datasets from the Internet has led to unmatched generalization for natural language understanding (NLU) tasks. However, such language models remain fragile when tasked with algorithmic forms of reasoning, where computations must be precise and robust. To address this limitation, we propose a novel approach that combines the Transformer's language understanding with the robustness of graph neural network (GNN)-based neural algorithmic reasoners (NARs). Such NARs proved effective as generic solvers for algorithmic tasks, when specified in graph form. To make their embeddings accessible to a Transformer, we propose a hybrid architecture with a two-phase training procedure, allowing the tokens in the language model to cross-attend to the node embeddings from the NAR. We evaluate our resulting TransNAR model on CLRS-Text, the text-based version of the CLRS-30 benchmark, and demonstrate significant gains over Transformer-only models for algorithmic reasoning, both in and out of distribution., Comment: To appear at CVPR 2024 Multimodal Algorithmic Reasoning (MAR) Workshop. 10 pages, 5 figures

In differentially private (DP) machine learning, the privacy guarantees of DP mechanisms are often reported and compared on the basis of a single $(\varepsilon, \delta)$-pair. This practice overlooks that DP guarantees can vary substantially even between mechanisms sharing a given $(\varepsilon, \delta)$, and potentially introduces privacy vulnerabilities which can remain undetected. This motivates the need for robust, rigorous methods for comparing DP guarantees in such cases. Here, we introduce the $\Delta$-divergence between mechanisms which quantifies the worst-case excess privacy vulnerability of choosing one mechanism over another in terms of $(\varepsilon, \delta)$, $f$-DP and in terms of a newly presented Bayesian interpretation. Moreover, as a generalisation of the Blackwell theorem, it is endowed with strong decision-theoretic foundations. Through application examples, we show that our techniques can facilitate informed decision-making and reveal gaps in the current understanding of privacy risks, as current practices in DP-SGD often result in choosing mechanisms with high excess privacy vulnerabilities., Comment: ICML 2024

The present work concerns the detailed construction of a holographic model for a type-II s-wave superconductor defined on a 5-dimensional anisotropic rotating black hole. We examine the role of rotation and anisotropy on the properties of the superconductor model focusing on the condensate and the AC conductivity, for which we obtain closed formulas, using both analytical and numerical methods. The results reveal that the rotation is responsible for the appearance of a peak and for introducing an exponentially vanishing behavior in the high-frequency limit of the real component of the AC conductivity. Such a behavior aligns with that observed in high-temperature superconductor models and experiments, where the peak and vanishing behavior result from quasiparticle damping, suggesting a plausible relation between the rotation of a black hole and quasiparticle damping effects in a superconducting material. In addition, we provide a detailed construction of the vortex lattice presented in arXiv:2208.05988 and study its behavior as a function of an external uniform magnetic field. Once again, it is shown that the vortex lattice can be continuously deformed along with a change in the vortex population by virtue of the magnetic field, providing a promising avenue for holographically modeling the vortex lattice deformations observed in experimental superconducting materials. As an observed experimental effect, we describe both the vortex lattice deformation and the vortex population increment under the action of an external magnetic field in the LiFeAs type-II superconductor. These effects supplement those previously found for the FeSe type-II superconductor studied in arXiv:2208.05988., Comment: 25 pages, 15 figures

Eliciting reasoning capabilities from language models (LMs) is a critical direction on the path towards building intelligent systems. Most recent studies dedicated to reasoning focus on out-of-distribution performance on procedurally-generated synthetic benchmarks, bespoke-built to evaluate specific skills only. This trend makes results hard to transfer across publications, slowing down progress. Three years ago, a similar issue was identified and rectified in the field of neural algorithmic reasoning, with the advent of the CLRS benchmark. CLRS is a dataset generator comprising graph execution traces of classical algorithms from the Introduction to Algorithms textbook. Inspired by this, we propose CLRS-Text -- a textual version of these algorithmic traces. Out of the box, CLRS-Text is capable of procedurally generating trace data for thirty diverse, challenging algorithmic tasks across any desirable input distribution, while offering a standard pipeline in which any additional algorithmic tasks may be created in the benchmark. We fine-tune and evaluate various LMs as generalist executors on this benchmark, validating prior work and revealing a novel, interesting challenge for the LM reasoning community. Our code is available at https://github.com/google-deepmind/clrs/tree/master/clrs/_src/clrs_text., Comment: Preprint, under review. Comments welcome

Computational marketing has become increasingly important in today's digital world, facing challenges such as massive heterogeneous data, multi-channel customer journeys, and limited marketing budgets. In this paper, we propose a general framework for marketing AI systems, the Neural Optimization with Adaptive Heuristics (NOAH) framework. NOAH is the first general framework for marketing optimization that considers both to-business (2B) and to-consumer (2C) products, as well as both owned and paid channels. We describe key modules of the NOAH framework, including prediction, optimization, and adaptive heuristics, providing examples for bidding and content optimization. We then detail the successful application of NOAH to LinkedIn's email marketing system, showcasing significant wins over the legacy ranking system. Additionally, we share details and insights that are broadly useful, particularly on: (i) addressing delayed feedback with lifetime value, (ii) performing large-scale linear programming with randomization, (iii) improving retrieval with audience expansion, (iv) reducing signal dilution in targeting tests, and (v) handling zero-inflated heavy-tail metrics in statistical testing., Comment: KDD 2024

The ever-increasing use of generative models in various fields where tabular data is used highlights the need for robust and standardized validation metrics to assess the similarity between real and synthetic data. Current methods lack a unified framework and rely on diverse and often inconclusive statistical measures. Divergences, which quantify discrepancies between data distributions, offer a promising avenue for validation. However, traditional approaches calculate divergences independently for each feature due to the complexity of joint distribution modeling. This paper addresses this challenge by proposing a novel approach that uses divergence estimation to overcome the limitations of marginal comparisons. Our core contribution lies in applying a divergence estimator to build a validation metric considering the joint distribution of real and synthetic data. We leverage a probabilistic classifier to approximate the density ratio between datasets, allowing the capture of complex relationships. We specifically calculate two divergences: the well-known Kullback-Leibler (KL) divergence and the Jensen-Shannon (JS) divergence. KL divergence offers an established use in the field, while JS divergence is symmetric and bounded, providing a reliable metric. The efficacy of this approach is demonstrated through a series of experiments with varying distribution complexities. The initial phase involves comparing estimated divergences with analytical solutions for simple distributions, setting a benchmark for accuracy. Finally, we validate our method on a real-world dataset and its corresponding synthetic counterpart, showcasing its effectiveness in practical applications. This research offers a significant contribution with applicability beyond tabular data and the potential to improve synthetic data validation in various fields., Comment: 15 pages, 14 figures

The chemical abundances of Milky Way's satellites reflect their star formation histories (SFHs), yet, due to the difficulty of determining the ages of old stars, the SFHs of most satellites are poorly measured. Ongoing and upcoming surveys will obtain around ten times more medium-resolution spectra for stars in satellites than are currently available. To correctly extract SFHs from large samples of chemical abundances, the relationship between chemical abundances and SFHs needs to be clarified. Here, we perform a high-resolution cosmological zoom-in simulation of a Milky Way-like galaxy with detailed models of star formation, supernova feedback, and metal diffusion. We quantify SFHs, metallicity distribution functions, and the $\alpha$-element (Mg, Ca, and Si) abundances in satellites of the host galaxy. We find that star formation in most simulated satellites is quenched before infalling to their host. Star formation episodes in simulated satellites are separated by a few hundred Myr owing to supernova feedback; each star formation event produces groups of stars with similar [$\alpha$/Fe] and [Fe/H]. We then perform a mock observation of the upcoming Subaru Prime Focus Spectrograph (PFS) observations. We find that Subaru PFS will be able to detect distinct groups of stars in [$\alpha$/Fe] vs. [Fe/H] space, produced by episodic star formation. This result means that episodic SFHs can be estimated from the chemical abundances of $\gtrsim$ 1,000 stars determined with medium-resolution spectroscopy., Comment: 16 pages, 9 figures, accepted for publication in The Astrophysical Journal

The growing use of large language model (LLM)-based conversational agents to manage sensitive user data raises significant privacy concerns. While these agents excel at understanding and acting on context, this capability can be exploited by malicious actors. We introduce a novel threat model where adversarial third-party apps manipulate the context of interaction to trick LLM-based agents into revealing private information not relevant to the task at hand. Grounded in the framework of contextual integrity, we introduce AirGapAgent, a privacy-conscious agent designed to prevent unintended data leakage by restricting the agent's access to only the data necessary for a specific task. Extensive experiments using Gemini, GPT, and Mistral models as agents validate our approach's effectiveness in mitigating this form of context hijacking while maintaining core agent functionality. For example, we show that a single-query context hijacking attack on a Gemini Ultra agent reduces its ability to protect user data from 94% to 45%, while an AirGapAgent achieves 97% protection, rendering the same attack ineffective., Comment: at CCS'24

We explore the practicality of early fault-tolerant quantum algorithms, focusing on ground-state energy estimation problems. Specifically, we address the computation of the cumulative distribution function (CDF) of the spectral measure of the Hamiltonian and the identification of its discontinuities. Scaling to bigger system sizes unveils three challenges: the smoothness of the CDF for large supports, the absence of tight lower bounds on the overlap with the actual ground state, and the complexity of preparing high-quality initial states. To tackle these challenges, we introduce a signal processing technique for identifying the inflection point of the CDF. We argue that this change of paradigm significantly simplifies the problem, making it more accessible while still being accurate. Hence, instead of trying to find the exact ground-state energy, we advocate improving on the classical estimate by aiming at the low-energy support of the initial state. Furthermore, we offer quantitative resource estimates for the maximum number of samples required to identify an increase in the CDF of a given size. Finally, we conduct numerical experiments on a 26-qubit fully-connected Heisenberg model using a truncated density-matrix renormalization group (DMRG) initial state of low bond dimension. Results show that the prediction obtained with the quantum algorithm aligns well with the DMRG-converged energy at large bond dimensions and requires several orders of magnitude fewer samples than predicted by the theory. Hence, we argue that CDF-based quantum algorithms are a viable, practical alternative to quantum phase estimation in resource-limited scenarios., Comment: 16 pages, 9 figures

This paper focuses on the opportunities and the ethical and societal risks posed by advanced AI assistants. We define advanced AI assistants as artificial agents with natural language interfaces, whose function is to plan and execute sequences of actions on behalf of a user, across one or more domains, in line with the user's expectations. The paper starts by considering the technology itself, providing an overview of AI assistants, their technical foundations and potential range of applications. It then explores questions around AI value alignment, well-being, safety and malicious uses. Extending the circle of inquiry further, we next consider the relationship between advanced AI assistants and individual users in more detail, exploring topics such as manipulation and persuasion, anthropomorphism, appropriate relationships, trust and privacy. With this analysis in place, we consider the deployment of advanced assistants at a societal scale, focusing on cooperation, equity and access, misinformation, economic impact, the environment and how best to evaluate advanced AI assistants. Finally, we conclude by providing a range of recommendations for researchers, developers, policymakers and public stakeholders.

Pediatric tumors of the central nervous system are the most common cause of cancer-related death in children. The five-year survival rate for high-grade gliomas in children is less than 20%. Due to their rarity, the diagnosis of these entities is often delayed, their treatment is mainly based on historic treatment concepts, and clinical trials require multi-institutional collaborations. Here we present the CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs challenge, focused on pediatric brain tumors with data acquired across multiple international consortia dedicated to pediatric neuro-oncology and clinical trials. The CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs challenge brings together clinicians and AI/imaging scientists to lead to faster development of automated segmentation techniques that could benefit clinical trials, and ultimately the care of children with brain tumors., Comment: arXiv admin note: substantial text overlap with arXiv:2305.17033

This paper studies the truncation method from Alquier [1] to derive high-probability PAC-Bayes bounds for unbounded losses with heavy tails. Assuming that the $p$-th moment is bounded, the resulting bounds interpolate between a slow rate $1 / \sqrt{n}$ when $p=2$, and a fast rate $1 / n$ when $p \to \infty$ and the loss is essentially bounded. Moreover, the paper derives a high-probability PAC-Bayes bound for losses with a bounded variance. This bound has an exponentially better dependence on the confidence parameter and the dependency measure than previous bounds in the literature. Finally, the paper extends all results to guarantees in expectation and single-draw PAC-Bayes. In order to so, it obtains analogues of the PAC-Bayes fast rate bound for bounded losses from [2] in these settings., Comment: 9 pages: 5 of main text, 1 of references, and 3 of appendices

A key aspect of intelligence is the ability to demonstrate a broad spectrum of behaviors for adapting to unexpected situations. Over the past decade, advancements in deep reinforcement learning have led to groundbreaking achievements to solve complex continuous control tasks. However, most approaches return only one solution specialized for a specific problem. We introduce Quality-Diversity Actor-Critic (QDAC), an off-policy actor-critic deep reinforcement learning algorithm that leverages a value function critic and a successor features critic to learn high-performing and diverse behaviors. In this framework, the actor optimizes an objective that seamlessly unifies both critics using constrained optimization to (1) maximize return, while (2) executing diverse skills. Compared with other Quality-Diversity methods, QDAC achieves significantly higher performance and more diverse behaviors on six challenging continuous control locomotion tasks. We also demonstrate that we can harness the learned skills to adapt better than other baselines to five perturbed environments. Finally, qualitative analyses showcase a range of remarkable behaviors: adaptive-intelligent-robotics.github.io/QDAC., Comment: The first two authors contributed equally to this work. Accepted at ICML 2024

This paper studies the Bayesian regret of a variant of the Thompson-Sampling algorithm for bandit problems. It builds upon the information-theoretic framework of [Russo and Van Roy, 2015] and, more specifically, on the rate-distortion analysis from [Dong and Van Roy, 2020], where they proved a bound with regret rate of $O(d\sqrt{T \log(T)})$ for the $d$-dimensional linear bandit setting. We focus on bandit problems with a metric action space and, using a chaining argument, we establish new bounds that depend on the metric entropy of the action space for a variant of Thompson-Sampling. Under suitable continuity assumption of the rewards, our bound offers a tight rate of $O(d\sqrt{T})$ for $d$-dimensional linear bandit problems., Comment: 15 pages: 8 of main text and 7 of appendices

The infrared (IR) range is extremely useful in the context of chemical abundance studies of the gas-phase interstellar medium (ISM) due to the large variety of ionic species traced in this regime, the negligible effects from dust attenuation or temperature stratification, and the amount of data that has been and will be released in the coming years. Taking advantage of available IR emission lines, we analysed the chemical content of the gas-phase ISM in a sample of 131 Star-Forming Galaxies (SFGs) and 73 Active Galactic Nuclei (AGNs). Particularly, we derived the chemical content via their total oxygen abundance in combination with nitrogen and sulfur abundances, and with the ionisation parameter. We used a new version of the code HII-CHI-Mistry-IR v3.1 which allows us to estimate log(N/O), 12+log(O/H), log(U), and, for the first time, 12+log(S/H) from IR emission lines, which can be applied to both SFGs and AGNs. We tested that the estimations from this new version, that only considers sulfur lines for the derivation of sulfur abundances, are compatible with previous studies. While most of the SFGs and AGNs show solar log(N/O) abundances, we found a large spread in the log(S/O) relative abundances. Specifically, we found extremely low log(S/O) values (1/10th solar) in some SFGs and AGNs with solar-like oxygen abundances. This result warns against the use of optical and IR sulfur emission lines to estimate oxygen abundances when no prior estimation of log(S/O) is provided., Comment: Accepted for publication in A&A. 16 pages, 11 figures, 4 electronic tables

We demonstrate an experimental protocol for the preparation and control of heralded single photons in inhomogeneously polarized states, such as Vector Vortex and Full Poincar\'e beam states. A laser beam is shaped by a voltage-controlled spin-to-orbital angular momentum converter q-plate device which eliminates the need for an interferometer for the robust preparation of high-quality inhomogeneously polarized beams. Such a beam is then used as pump in a spontaneous parametric downconversion (SPDC) photon-pair source. We demonstrate the full pump to heralded single photon transfer of the intensity/phase distributions, as well as of the vector polarization structure. Additionally, we show that by controlling the polarization to which the heralding idler photon is projected before detection, we can toggle between the direct and basis-switched pump-single photon transfer. We show that this non-local control of the heralded single photon pertains also to the topological class of the resulting heralded single photon. We believe that our work will lead to new opportunities in photons-based quantum information processing science., Comment: 11 pages, 5 figures

Experimentation is an intrinsic part of research in artificial intelligence since it allows for collecting quantitative observations, validating hypotheses, and providing evidence for their reformulation. For that reason, experimentation must be coherent with the purposes of the research, properly addressing the relevant questions in each case. Unfortunately, the literature is full of works whose experimentation is neither rigorous nor convincing, oftentimes designed to support prior beliefs rather than answering the relevant research questions. In this paper, we focus on the field of metaheuristic optimization, since it is our main field of work, and it is where we have observed the misconduct that has motivated this letter. Even if we limit the focus of this manuscript to the experimental part of the research, our main goal is to sew the seed of sincere critical assessment of our work, sparking a reflection process both at the individual and the community level. Such a reflection process is too complex and extensive to be tackled as a whole. Therefore, to bring our feet to the ground, we will include in this document our reflections about the role of experimentation in our work, discussing topics such as the use of benchmark instances vs instance generators, or the statistical assessment of empirical results. That is, all the statements included in this document are personal views and opinions, which can be shared by others or not. Certainly, having different points of view is the basis to establish a good discussion process.

Report Noisy Max and Above Threshold are two classical differentially private (DP) selection mechanisms. Their output is obtained by adding noise to a sequence of low-sensitivity queries and reporting the identity of the query whose (noisy) answer satisfies a certain condition. Pure DP guarantees for these mechanisms are easy to obtain when Laplace noise is added to the queries. On the other hand, when instantiated using Gaussian noise, standard analyses only yield approximate DP guarantees despite the fact that the outputs of these mechanisms lie in a discrete space. In this work, we revisit the analysis of Report Noisy Max and Above Threshold with Gaussian noise and show that, under the additional assumption that the underlying queries are bounded, it is possible to provide pure ex-ante DP bounds for Report Noisy Max and pure ex-post DP bounds for Above Threshold. The resulting bounds are tight and depend on closed-form expressions that can be numerically evaluated using standard methods. Empirically we find these lead to tighter privacy accounting in the high privacy, low data regime. Further, we propose a simple privacy filter for composing pure ex-post DP guarantees, and use it to derive a fully adaptive Gaussian Sparse Vector Technique mechanism. Finally, we provide experiments on mobility and energy consumption datasets demonstrating that our Sparse Vector Technique is practically competitive with previous approaches and requires less hyper-parameter tuning., Comment: AISTATS 2024

Cities play a vital role in providing water to vast populations. Due to an increasing population and the urbanization process, urban water demand is expected to double by 2050. However, some cities are expanding in locations with limited water retention capacity, scarce rainfall and unfavourable shape. This study quantifies the impact of urban form on water scarcity across more than 100 cities in Latin America, Asia, and Africa. The analysis integrates indicators of urban morphology, satellite imagery for natural resources, and terrain metrics to model cities' influence on water accessibility. Water tariffs, proximity to critical infrastructure, and access to piped water are used to assess scarcity in urban areas. Results show that locations that are far from the centre, with smaller buildings and fewer constructed surfaces, are less likely to have water access. Urban sprawl negatively impacts water availability within a city, decreases proximity to critical infrastructure and increases water tariffs. Keeping everything else constant, a location that is twice as far from the city centre has 38% less proximity to critical infrastructure. Also, water tariffs are up to 75% higher and the capacity of cities to provide water drops by half in more sparse cities., Comment: 20 pages, 10 figures

Topological quantum error correction is a milestone in the scaling roadmap of quantum computers, which targets circuits with trillions of gates that would allow running quantum algorithms for real-world problems. The square-lattice surface code has become the workhorse to address this challenge, as it poses milder requirements on current devices both in terms of required error rates and small local connectivities. In some platforms, however, the connectivities are kept even lower in order to minimise gate errors at the hardware level, which limits the error correcting codes that can be directly implemented on them. In this work, we make a comparative study of possible strategies to overcome this limitation for the heavy-hexagonal lattice, the architecture of current IBM superconducting quantum computers. We explore two complementary strategies: the search for an efficient embedding of the surface code into the heavy-hexagonal lattice, as well as the use of codes whose connectivity requirements are naturally tailored to this architecture, such as subsystem-type and Floquet codes. Using noise models of increased complexity, we assess the performance of these strategies for IBM devices in terms of their error thresholds and qubit footprints. An optimized SWAP-based embedding of the surface code is found to be the most promising strategy towards a near-term demonstration of quantum error correction advantage.

Visible light communication (VLC) is a technology that complements radio frequency (RF) to fulfill the ever-increasing demand for wireless data traffic. The ubiquity of light-emitting diodes (LEDs), exploited as transmitters, increases the VLC market penetration and positions it as one of the most promising technologies to alleviate the spectrum scarcity of RF. However, VLC deployment is hindered by blockage causing connectivity outages in the presence of obstacles. Recently, optical reconfigurable intelligent surfaces (ORISs) have been considered to mitigate this problem. While prior works exploit ORISs for data or secrecy rate maximization, this paper studies the optimal placement of mirrors and ORISs, and the LED power allocation, for jointly minimizing the outage probability while keeping the lighting standards. We describe an optimal outage minimization framework and present solvable heuristics. We provide extensive numerical results and show that the use of ORISs may reduce the outage probability by up to 67% with respect to a no-mirror scenario and provide a gain of hundreds of kbit/J in optical energy efficiency with respect to the presented benchmark.

Stochastic Gradient Descent (SGD) stands as a cornerstone optimization algorithm with proven real-world empirical successes but relatively limited theoretical understanding. Recent research has illuminated a key factor contributing to its practical efficacy: the implicit regularization it instigates. Several studies have investigated the linear stability property of SGD in the vicinity of a stationary point as a predictive proxy for sharpness and generalization error in overparameterized neural networks (Wu et al., 2022; Jastrzebski et al., 2019; Cohen et al., 2021). In this paper, we delve deeper into the relationship between linear stability and sharpness. More specifically, we meticulously delineate the necessary and sufficient conditions for linear stability, contingent on hyperparameters of SGD and the sharpness at the optimum. Towards this end, we introduce a novel coherence measure of the loss Hessian that encapsulates pertinent geometric properties of the loss function that are relevant to the linear stability of SGD. It enables us to provide a simplified sufficient condition for identifying linear instability at an optimum. Notably, compared to previous works, our analysis relies on significantly milder assumptions and is applicable for a broader class of loss functions than known before, encompassing not only mean-squared error but also cross-entropy loss., Comment: To appear at ICLR 2024

Network time series are becoming increasingly relevant in the study of dynamic processes characterised by a known or inferred underlying network structure. Generalised Network Autoregressive (GNAR) models provide a parsimonious framework for exploiting the underlying network, even in the high-dimensional setting. We extend the GNAR framework by presenting the $\textit{community}$-$\alpha$ GNAR model that exploits prior knowledge and/or exogenous variables for identifying and modelling dynamic interactions across communities in the network. We further analyse the dynamics of $\textit{ Red, Blue}$ and $\textit{Swing}$ states throughout presidential elections in the USA. Our analysis suggests interesting global and communal effects., Comment: 21 pages, 12 figures. Preprint IFCS 2024

Aims. We aim to expand the atmospheric exploration of AF Lep b by modeling all available observations obtained with SPHERE at VLT (between 0.95-1.65, at 2.105, and 2.253 $\mu$m, and NIRC2 at Keck (at 3.8 $\mu$m) with self-consistent atmospheric models. Methods. To understand the physical properties of this exoplanet, we used ForMoSA. This forward-modeling code compares observations with grids of pre-computed synthetic atmospheric spectra using Bayesian inference methods. We used Exo-REM, an atmospheric radiative-convective equilibrium model, including the effects of non-equilibrium processes and clouds. Results. From the atmospheric modeling we derive solutions at a low effective temperature of ~750 K. Our analysis also favors a metal-rich atmosphere (>0.4) and solar to super-solar carbon-to-oxygen ratio (~0.6). We tested the robustness of the estimated values for each parameter by cross-validating our models using the leave-one-out strategy, where all points are used iteratively as validation points. Our results indicate that the photometry point at 3.8 $\mu$m strongly drives the metal-rich and super-solar carbon-to-oxygen solutions. Conclusions. Our atmospheric forward-modeling analysis strongly supports the planetary nature of AF Lep b. Its spectral energy distribution is consistent with that of a young, cold, early-T super-Jovian planet. We recover physically consistent solutions for the surface gravity and radius, which allows us to reconcile atmospheric forward modeling with evolutionary models, in agreement with the previously published complementary analysis done by retrievals. Finally, we identified that future data at longer wavelengths are mandatory before concluding about the metal-rich nature of AF Lep b., Comment: 7 pages, 7 figures, 6 tables. Accepted for publication on the 9th of January 2024 in A&A

In this paper we show how tensor networks help in developing explainability of machine learning algorithms. Specifically, we develop an unsupervised clustering algorithm based on Matrix Product States (MPS) and apply it in the context of a real use-case of adversary-generated threat intelligence. Our investigation proves that MPS rival traditional deep learning models such as autoencoders and GANs in terms of performance, while providing much richer model interpretability. Our approach naturally facilitates the extraction of feature-wise probabilities, Von Neumann Entropy, and mutual information, offering a compelling narrative for classification of anomalies and fostering an unprecedented level of transparency and interpretability, something fundamental to understand the rationale behind artificial intelligence decisions., Comment: 9 pages, 9 figures, 2 table, minor typos corrected

Network time series are becoming increasingly important across many areas in science and medicine and are often characterised by a known or inferred underlying network structure, which can be exploited to make sense of dynamic phenomena that are often high-dimensional. For example, the Generalised Network Autoregressive (GNAR) models exploit such structure parsimoniously. We use the GNAR framework to introduce two association measures: the network and partial network autocorrelation functions, and introduce Corbit (correlation-orbit) plots for visualisation. As with regular autocorrelation plots, Corbit plots permit interpretation of underlying correlation structures and, crucially, aid model selection more rapidly than using other tools such as AIC or BIC. We additionally interpret GNAR processes as generalised graphical models, which constrain the processes' autoregressive structure and exhibit interesting theoretical connections to graphical models via utilization of higher-order interactions. We demonstrate how incorporation of prior information is related to performing variable selection and shrinkage in the GNAR context. We illustrate the usefulness of the GNAR formulation, network autocorrelations and Corbit plots by modelling a COVID-19 network time series of the number of admissions to mechanical ventilation beds at 140 NHS Trusts in England & Wales. We introduce the Wagner plot that can analyse correlations over different time periods or with respect to external covariates. In addition, we introduce plots that quantify the relevance and influence of individual nodes. Our modelling provides insight on the underlying dynamics of the COVID-19 series, highlights two groups of geographically co-located `influential' NHS Trusts and demonstrates superior prediction abilities when compared to existing techniques.

This paper addresses the trajectory-tracking problem for a class of electromechanical systems. To this end, the dynamics of the plants are modeled in the so-called port-Hamiltonian framework. Then, the notion of contraction is exploited to design the desired closed-loop dynamics and the corresponding tracking controller. Notably, the proposed control design method does not require solving partial differential equations or changing the coordinates of the plant, which permits preserving the physical interpretation of the controller. The applicability of the proposed approach is illustrated in several electromechanical systems via simulations.

Here we introduce an improved approach to Variational Quantum Attack Algorithms (VQAA) on crytographic protocols. Our methods provide robust quantum attacks to well-known cryptographic algorithms, more efficiently and with remarkably fewer qubits than previous approaches. We implement simulations of our attacks for symmetric-key protocols such as S-DES, S-AES and Blowfish. For instance, we show how our attack allows a classical simulation of a small 8-qubit quantum computer to find the secret key of one 32-bit Blowfish instance with 24 times fewer number of iterations than a brute-force attack. Our work also shows improvements in attack success rates for lightweight ciphers such as S-DES and S-AES. Further applications beyond symmetric-key cryptography are also discussed, including asymmetric-key protocols and hash functions. In addition, we also comment on potential future improvements of our methods. Our results bring one step closer assessing the vulnerability of large-size classical cryptographic protocols with Noisy Intermediate-Scale Quantum (NISQ) devices, and set the stage for future research in quantum cybersecurity., Comment: 12 pages, 8 figures

The construction industry has been notoriously slow to adopt new technology and embrace automation. This has resulted in lower efficiency and productivity compared to other industries where automation has been widely adopted. However, recent advancements in robotics and artificial intelligence offer a potential solution to this problem. In this study, a methodology is proposed to integrate multi-robotic systems in construction projects with the aim of increasing efficiency and productivity. The proposed approach involves the use of multiple robot and human agents working collaboratively to complete a construction task. The methodology was tested through a case study that involved 3D digitization of a small, occluded space using two robots and one human agent. The results show that integrating multi-agent robotic systems in construction can effectively overcome challenges and complete tasks efficiently. The implications of this study suggest that multi-agent robotic systems could revolutionize the industry.

Malware authors have seen obfuscation as the mean to bypass malware detectors based on static analysis features. For Android, several studies have confirmed that many anti-malware products are easily evaded with simple program transformations. As opposed to these works, ML detection proposals for Android leveraging static analysis features have also been proposed as obfuscation-resilient. Therefore, it needs to be determined to what extent the use of a specific obfuscation strategy or tool poses a risk for the validity of ML malware detectors for Android based on static analysis features. To shed some light in this regard, in this article we assess the impact of specific obfuscation techniques on common features extracted using static analysis and determine whether the changes are significant enough to undermine the effectiveness of ML malware detectors that rely on these features. The experimental results suggest that obfuscation techniques affect all static analysis features to varying degrees across different tools. However, certain features retain their validity for ML malware detection even in the presence of obfuscation. Based on these findings, we propose a ML malware detector for Android that is robust against obfuscation and outperforms current state-of-the-art detectors.

We present an implementation of a Web3 platform that leverages the Groth16 Zero-Knowledge Proof schema to verify the validity of questionnaire results within Smart Contracts. Our approach ensures that the answer key of the questionnaire remains undisclosed throughout the verification process, while ensuring that the evaluation is done fairly. To accomplish this, users respond to a series of questions, and their answers are encoded and securely transmitted to a hidden backend. The backend then performs an evaluation of the user's answers, generating the overall result of the questionnaire. Additionally, it generates a Zero-Knowledge Proof, attesting that the answers were appropriately evaluated against a valid set of constraints. Next, the user submits their result along with the proof to a Smart Contract, which verifies their validity and issues a non-fungible token (NFT) as an attestation of the user's test result. In this research, we implemented the Zero-Knowledge functionality using Circom 2 and deployed the Smart Contract using Solidity, thereby showcasing a practical and secure solution for questionnaire validity verification in the context of Smart Contracts., Comment: 5 pages, 2 figures, presented at CISMA 2023 Guanajuato, Mexico

Daniyar Shamkanov proved that three distinct systems of sequent calculi for GL are equivalent. These systems consist in one with finite proofs, another with ill-founded proofs and the last one with cyclic proofs. The main tool used for proving the equivalence is corecursion. In this project, we prove the equivalence between a finitary sequent calculus for iGL and a cyclic calculus, using also coinductive methods.

We introduce the SDSS/APOGEE Value Added Catalogue of Galactic Globular Cluster (GC) Stars. The catalogue is the result of a critical search of the APOGEE data release 17 (DR17) catalogue for candidate members of all known Galactic GCs. Candidate members are assigned to various GCs on the basis of position on the sky, proper motion, and radial velocity. The catalogue contains a total of 7,737 entries for 6,422 unique stars associated with 72 Galactic GCs. Full APOGEE DR17 information is provided, including radial velocities and abundances for up to 20 elements. Membership probabilities estimated on the basis of precision radial velocities are made available. Comparisons with chemical compositions derived by the GALAH survey, as well as optical values from the literature, show good agreement. This catalogue represents a significant increase in the public database of GC star chemical compositions and kinematics, providing a massive homogeneous data set that will enable a variety of studies. The catalogue in fits format is available for public download from the SDSS-IV DR17 value added catalogue website., Comment: To appear in MNRAS. The paper is 16 pages long, containing 16 figures, 3 tables, and 1 Appendix. The catalogue and metadata can be obtained upon request to the authors

In this article we examine recent developments in the research area concerning the creation of end-to-end models for the complete optimization of measuring instruments. The models we consider rely on differentiable programming methods and on the specification of a software pipeline including all factors impacting performance -- from the data-generating processes to their reconstruction and the extraction of inference on the parameters of interest of a measuring instrument -- along with the careful specification of a utility function well aligned with the end goals of the experiment. Building on previous studies originated within the MODE Collaboration, we focus specifically on applications involving instruments for particle physics experimentation, as well as industrial and medical applications that share the detection of radiation as their data-generating mechanism., Comment: 70 pages, 17 figures. To be submitted to journal

We revisit the classical $K_Z$ problem -- determination of the vertical force and implied total mass density distribution of the Milky Way disk -- for a wide range of Galactocentric radius and vertical height using chemically selected thin and thick disk samples based on APOGEE spectroscopy combined with the Gaia astrometry. We derived the velocity dispersion profiles in Galactic cylindrical coordinates, and solved the Jeans Equation for the two samples separately. The result is surprising that the total surface mass density as a function of vertical height as derived for these two chemically distinguished populations are different. The discrepancies are larger in the inner compared to the outer Galaxy, with the density calculated from thick disk being larger, independent of the Galactic radius. Furthermore, while there is an overall good agreement between the total mass density derived for the thick disk population and the Standard Halo Model for vertical heights larger than 1 kpc, close to the midplane the mass density observed using the thick disk population is larger than the predicted from the Standard Halo Model. We explore various implications of these discrepancies, and speculate their sources, including problems associated with the assumed density laws, velocity dispersion profiles, and the Galactic rotation curve, potential non-equilibrium of the Galactic disk, or a failure of the NFW dark matter halo profile for the Milky Way. We conclude that the growing detail in hand on the chemodynamical distributions of Milky Way stars challenges traditional analytical treatments of the $K_Z$ problem., Comment: Accepted by MNRAS

Privacy-preserving machine learning aims to train models on private data without leaking sensitive information. Differential privacy (DP) is considered the gold standard framework for privacy-preserving training, as it provides formal privacy guarantees. However, compared to their non-private counterparts, models trained with DP often have significantly reduced accuracy. Private classifiers are also believed to exhibit larger performance disparities across subpopulations, raising fairness concerns. The poor performance of classifiers trained with DP has prevented the widespread adoption of privacy preserving machine learning in industry. Here we show that pre-trained foundation models fine-tuned with DP can achieve similar accuracy to non-private classifiers, even in the presence of significant distribution shifts between pre-training data and downstream tasks. We achieve private accuracies within a few percent of the non-private state of the art across four datasets, including two medical imaging benchmarks. Furthermore, our private medical classifiers do not exhibit larger performance disparities across demographic groups than non-private models. This milestone to make DP training a practical and reliable technology has the potential to widely enable machine learning practitioners to train safely on sensitive datasets while protecting individuals' privacy.

The mechanisms behind the success of multi-view self-supervised learning (MVSSL) are not yet fully understood. Contrastive MVSSL methods have been studied through the lens of InfoNCE, a lower bound of the Mutual Information (MI). However, the relation between other MVSSL methods and MI remains unclear. We consider a different lower bound on the MI consisting of an entropy and a reconstruction term (ER), and analyze the main MVSSL families through its lens. Through this ER bound, we show that clustering-based methods such as DeepCluster and SwAV maximize the MI. We also re-interpret the mechanisms of distillation-based approaches such as BYOL and DINO, showing that they explicitly maximize the reconstruction term and implicitly encourage a stable entropy, and we confirm this empirically. We show that replacing the objectives of common MVSSL methods with this ER bound achieves competitive performance, while making them stable when training with smaller batch sizes or smaller exponential moving average (EMA) coefficients. Github repo: https://github.com/apple/ml-entropy-reconstruction., Comment: 18 pages: 9 of main text, 2 of references, and 7 of supplementary material [Updated typo in page 6 (Section 3.2)]. Appears in the proceedings of ICML 2023

The build up of heavy elements and the stellar mass assembly are fundamental processes in the formation and evolution of galaxies. Although they have been extensively studied through observations and simulations, the key elements that govern these processes, such as gas accretion and outflows, are not fully understood. This is especially true for luminous and massive galaxies, which usually suffer strong feedback in the form of massive outflows, and large-scale gas accretion triggered by galaxy interactions. For a sample of 77 luminous infrared (IR) galaxies, we derive chemical abundances using new diagnostics based on nebular IR lines, which peer through the dusty medium of these objects and allow us to include the obscured metals in our abundance determinations. In contrast to optical-based studies, our analysis reveals that most luminous IR galaxies remain close to the mass-metallicity relation. Nevertheless, four galaxies with extreme star-formation rates ($> 60$M$_{\odot }$yr$^{-1}$) in their late merger stages show heavily depressed metallicities of 12+log(O/H) $\sim 7.7$--$8.1$ along with solar-like N/O ratios, indicative of gas mixing processes affecting their chemical composition. This evidence suggests the action of a massive infall of metal-poor gas in a short phase during the late merger stages, eventually followed by a rapid enrichment. These results challenge the classical gas equilibrium scenario usually applied to main-sequence galaxies, suggesting that the chemical enrichment and stellar-mass growth in luminous IR galaxies are regulated by different processes., Comment: Submitted to Nature Astronomy. 19 pages, 4 figures

As our nearest large neighbor, the Andromeda Galaxy provides a unique laboratory for investigating galaxy formation and the distribution and substructure properties of dark matter in a Milky Way-like galaxy. Here, we propose an initial 2-epoch ($\Delta t\approx 5$yr), 2-band Roman survey of the entire halo of Andromeda, covering 500 square degrees, which will detect nearly every red giant star in the halo (10$\sigma$ detection in F146, F062 of 26.5, 26.1AB mag respectively) and yield proper motions to $\sim$25 microarcsec/year (i.e., $\sim$90 km/s) for all stars brighter than F146 $\approx 23.6$ AB mag (i.e., reaching the red clump stars in the Andromeda halo). This survey will yield (through averaging) high-fidelity proper motions for all satellites and compact substructures in the Andromeda halo and will enable statistical searches for clusters in chemo-dynamical space. Adding a third epoch during the extended mission will improve these proper motions by $\sim t^{-1.5}$, to $\approx 11$ km/s, but this requires obtaining the first epoch in Year 1 of Roman operations. In combination with ongoing and imminent spectroscopic campaigns with ground-based telescopes, this Roman survey has the potential to yield full 3-d space motions of $>$100,000 stars in the Andromeda halo, including (by combining individual measurements) robust space motions of its entire globular cluster and most of its dwarf galaxy satellite populations. It will also identify high-velocity stars in Andromeda, providing unique information on the processes that create this population. These data offer a unique opportunity to study the immigration history, halo formation, and underlying dark matter scaffolding of a galaxy other than our own., Comment: Submitted in response to the call for Roman Space Telescope Core Community Survey white papers

In this paper, we present new high-probability PAC-Bayes bounds for different types of losses. Firstly, for losses with a bounded range, we recover a strengthened version of Catoni's bound that holds uniformly for all parameter values. This leads to new fast-rate and mixed-rate bounds that are interpretable and tighter than previous bounds in the literature. In particular, the fast-rate bound is equivalent to the Seeger--Langford bound. Secondly, for losses with more general tail behaviors, we introduce two new parameter-free bounds: a PAC-Bayes Chernoff analogue when the loss' cumulative generating function is bounded, and a bound when the loss' second moment is bounded. These two bounds are obtained using a new technique based on a discretization of the space of possible events for the ``in probability'' parameter optimization problem. This technique is both simpler and more general than previous approaches optimizing over a grid on the parameters' space. Finally, using a simple technique that is applicable to any existing bound, we extend all previous results to anytime-valid bounds., Comment: 43 pages: ~20 of main text, ~6.5 of references, and ~17.5 of appendices. Published at JMLR

In this paper, we study the approximate minimization problem of weighted finite automata (WFAs): to compute the best possible approximation of a WFA given a bound on the number of states. By reformulating the problem in terms of Hankel matrices, we leverage classical results on the approximation of Hankel operators, namely the celebrated Adamyan-Arov-Krein (AAK) theory. We solve the optimal spectral-norm approximate minimization problem for irredundant WFAs with real weights, defined over a one-letter alphabet. We present a theoretical analysis based on AAK theory, and bounds on the quality of the approximation in the spectral norm and $\ell^2$ norm. Moreover, we provide a closed-form solution, and an algorithm, to compute the optimal approximation of a given size in polynomial time., Comment: 32 pages. arXiv admin note: substantial text overlap with arXiv:2102.06860

Squeezed light is injected into the dark port of gravitational wave interferometers, in order to reduce the quantum noise. A fraction of the interferometer output light can reach the OPO due to sub-optimal isolation of the squeezing injection path. This backscattered light interacts with squeezed light generation process, introducing additional measurement noise. We present a theoretical description of the noise coupling mechanism. We propose a control scheme to achieve a de-amplification of the backscattered light inside the OPO with a consequent reduction of the noise caused by it. The scheme was implemented at the GEO 600 detector and has proven to be crucial in maintaining a good level of quantum noise reduction of the interferometer for high parametric gain of the OPO. In particular, the mitigation of the backscattered light noise helped in reaching 6dB of quantum noise reduction [Phys. Rev. Lett. 126, 041102 (2021)]. The impact of backscattered-light-induced noise on the squeezing performance is phenomenologically equivalent to increased phase noise of the squeezing angle control. The results discussed in this paper provide a way for a more accurate estimation of the residual phase noise of the squeezed light field., Comment: 14 pages, 6 figures

Pediatric tumors of the central nervous system are the most common cause of cancer-related death in children. The five-year survival rate for high-grade gliomas in children is less than 20\%. Due to their rarity, the diagnosis of these entities is often delayed, their treatment is mainly based on historic treatment concepts, and clinical trials require multi-institutional collaborations. The MICCAI Brain Tumor Segmentation (BraTS) Challenge is a landmark community benchmark event with a successful history of 12 years of resource creation for the segmentation and analysis of adult glioma. Here we present the CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs 2023 challenge, which represents the first BraTS challenge focused on pediatric brain tumors with data acquired across multiple international consortia dedicated to pediatric neuro-oncology and clinical trials. The BraTS-PEDs 2023 challenge focuses on benchmarking the development of volumentric segmentation algorithms for pediatric brain glioma through standardized quantitative performance evaluation metrics utilized across the BraTS 2023 cluster of challenges. Models gaining knowledge from the BraTS-PEDs multi-parametric structural MRI (mpMRI) training data will be evaluated on separate validation and unseen test mpMRI dataof high-grade pediatric glioma. The CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs 2023 challenge brings together clinicians and AI/imaging scientists to lead to faster development of automated segmentation techniques that could benefit clinical trials, and ultimately the care of children with brain tumors.