Search

Optical systems that combine nonlinearity with coupling between various subsystems offer a flexible platform for observing a diverse range of nonlinear dynamics. Furthermore, engineering tolerances are such that the subsystems can be identical to within a fraction of the wavelength of light; hence, such coupled systems inherently have a natural symmetry that can lead to either delocalization or symmetry breaking. We consider here an optical Kerr cavity that supports two interacting electric fields, generated by two symmetric input beams. Mathematically, this system is modeled by a four-dimensional $\mathbb{Z}_2$-equivariant vector field with the strength and detuning of the input light as control parameters. Previous research has shown that complex switching dynamics are observed both experimentally and numerically across a wide range of parameter values. Here, we show that particular switching patterns are created at specific global bifurcations through either delocalization or symmetry breaking of a chaotic attractor. We find that the system exhibits infinitely many of these global bifurcations, which are organized by $\mathbb{Z}_2$-equivariant codimension-two Belyakov transitions. We investigate these switching dynamics by means of the continuation of global bifurcations in combination with the computation of kneading invariants and Lyapunov exponents. In this way, we provide a comprehensive picture of the interplay between different switching patterns of periodic orbits and chaotic attractors.

Early warning signals (EWSs) for complex dynamical systems aim to anticipate tipping points, or large regime shifts, before they occur. While signals computed from time series data, such as temporal variance and lagged autocorrelation functions, are useful for this task, they are costly to obtain in practice because they need many samples over time to calculate. Spatial EWSs use just a single sample per spatial location and aggregate the samples over space rather than time to try to mitigate this limitation. However, although many complex systems in nature and society form diverse networks, the performance of spatial EWSs is mostly unknown for general networks because the vast majority of studies of spatial EWSs have been on regular lattice networks. Therefore, we have carried out a comprehensive investigation of four major spatial EWSs on various networks. We find that the winning EWS depends on tipping scenarios, while spatial skewness tends to excel when tipping occurs via a saddle-node bifurcation, which is a commonly studied scenario in the literature. We also find that spatial EWSs behave in a drastically different manner between the square lattice and complex networks and tend to be more reliable for the latter than the former according to a standard performance measure. The present results encourage further studies of spatial EWSs on complex networks.

The emergence of anomalous Hall effects (AHE) in antiferromagnets presents an intriguing phenomenon with potential spintronic device applications due to their ultrafast switching dynamics. Mn3NiN antiperovskite stands out as a promising candidate owing to its noncollinear antiferromagnetism. In this study, we report a significant AHE observed in epitaxial Mn3NiN thin films below the N\'eel temperature, comparable to those observed in typical ferromagnets. Through a combination of magnetometry, neutron diffraction and ab-initio data analyses, we propose that the substantial AHE can be attributed to the non-vanishing Berry curvature arising from the {\Gamma}4g-type spin structure, coupled with uncompensated magnetic domains featuring a weak canted moment. This coupling, due to robust antiferromagnetic domains, enables deterministic detection and switching of AHE via the application of a magnetic field.

We report on the polarization dynamics of regenerative light pulses in a micropillar laser with integrated saturable absorber coupled to an external feedback mirror. The delayed self-coupled microlaser is operated in the excitable regime, where it regenerates incident pulses with a supra-threshold intensity -- resulting in a pulse train with inter-pulse period approximately given by the feedback delay time, in analogy with a self-coupled biological neuron. We report the experimental observation of vectorial breathers in polarization angle, manifesting themselves as a modulation of the linear polarized intensity components without significant modulation of the total intensity. Numerical analysis of a suitable model reveals that the observed polarization mode competition is a consequence of symmetry-breaking bifurcations induced by polarization anisotropy. Our model reproduces well the observed experimental results and predicts different regimes as a function of the polarization anisotropy parameters and the pump parameter. We believe that these findings are relevant for the fabrication of flexible sources of polarized pulses with controlled properties, as well as for neuroinspired on-chip computing applications, where the polarization may be used to encode or process information in novel ways., Comment: 5 pages, 4 figures

With the growing integration of AI in daily life, ensuring the robustness of systems to inference-time attacks is crucial. Among the approaches for certifying robustness to such adversarial examples, randomized smoothing has emerged as highly promising due to its nature as a wrapper around arbitrary black-box models. Previous work on randomized smoothing in natural language processing has primarily focused on specific subsets of edit distance operations, such as synonym substitution or word insertion, without exploring the certification of all edit operations. In this paper, we adapt Randomized Deletion (Huang et al., 2023) and propose, CERTified Edit Distance defense (CERT-ED) for natural language classification. Through comprehensive experiments, we demonstrate that CERT-ED outperforms the existing Hamming distance method RanMASK (Zeng et al., 2023) in 4 out of 5 datasets in terms of both accuracy and the cardinality of the certificate. By covering various threat models, including 5 direct and 5 transfer attacks, our method improves empirical robustness in 38 out of 50 settings., Comment: 22 pages, 3 figures, 12 tables. Include 11 pages of appendices

A fundamental premise of statistical physics is that the particles in a physical system are interchangeable, and hence the state of each specific component is representative of the system as a whole. This assumption breaks down for complex networks, in which nodes may be extremely diverse, and no single component can truly represent the state of the entire system. It seems, therefore, that to observe the dynamics of social, biological or technological networks, one must extract the dynamic states of a large number of nodes -- a task that is often practically prohibitive. To overcome this challenge, we use machine learning techniques to detect the network's sentinel nodes, a set of network components whose combined states can help approximate the average dynamics of the entire network. The method allows us to assess the state of a large complex system by tracking just a small number of carefully selected nodes. The resulting sentinel node set offers a natural probe by which to practically observe complex network dynamics.

Reuse of data in adaptive workflows poses challenges regarding overfitting and the statistical validity of results. Previous work has demonstrated that interacting with data via differentially private algorithms can mitigate overfitting, achieving worst-case generalization guarantees with asymptotically optimal data requirements. However, such past work assumes data is static and cannot accommodate situations where data grows over time. In this paper we address this gap, presenting the first generalization bounds for adaptive analysis in the dynamic data setting. We allow the analyst to adaptively schedule their queries conditioned on the current size of the data, in addition to previous queries and responses. We also incorporate time-varying empirical accuracy bounds and mechanisms, allowing for tighter guarantees as data accumulates. In a batched query setting, the asymptotic data requirements of our bound grows with the square-root of the number of adaptive queries, matching prior works' improvement over data splitting for the static setting. We instantiate our bound for statistical queries with the clipped Gaussian mechanism, where it empirically outperforms baselines composed from static bounds.

We focus on the existence and persistence of families of saddle periodic orbits in a four-dimensional Hamiltonian reversible ordinary differential equation derived using a travelling wave ansatz from a generalised nonlinear Schr{\"o}dinger equation (GNLSE) with quartic dispersion. In this way, we are able to characterise different saddle periodic orbits with different signatures that serve as organising centres of homoclinic orbits in the ODE and solitons in the GNLSE. To achieve our objectives, we employ numerical continuation techniques to compute these saddle periodic orbits, and study how they organise themselves as surfaces in phase space that undergo changes as a single parameter is varied. Notably, different surfaces of saddle periodic orbits can interact with each other through bifurcations that can drastically change their overall geometry or even create new surfaces of periodic orbits. Particularly we identify three different bifurcations: symmetry-breaking, period-$k$ multiplying, and saddle-node bifurcations. Each bifurcation exhibits a degenerate case, which subsequently gives rise to two bifurcations of the same type that occurs at particular energy levels that vary as a parameter is gradually increased. Additionally, we demonstrate how these degenerate bifurcations induce structural changes in the periodic orbits that can support homoclinic orbits by computing sequences of period-$k$ multiplying bifurcations.

Human collective tasks in teams and organizations increasingly require participation of members with diverse backgrounds working in networked social environments. However, little is known about how network structure and the functional diversity of member backgrounds would interact with each other and affect collective processes. Here we conducted three sets of human-subject experiments which involved 617 university students who collaborated anonymously in a collective ideation task on a custom-made online social network platform. We found that spatially clustered collectives with assortative background distribution tended to explore more diverse ideas than in other conditions, whereas collectives with random background distribution consistently generated ideas with the highest utility. We also found that higher network connectivity may improve individuals' overall experience but may not improve the collective performance regarding idea generation, idea diversity, and final idea quality., Comment: 50 pages, 23 figures, 3 tables

Microresonators are micron-scale optical systems that confine light using total internal reflection. These optical systems have gained interest in the last two decades due to their compact sizes, unprecedented measurement capabilities, and widespread applications. The increasingly high finesse (or $Q$ factor) of such resonators means that nonlinear effects are unavoidable even for low power, making them attractive for nonlinear applications, including optical comb generation and second harmonic generation. In addition, light in these nonlinear resonators may exhibit chaotic behavior across wide parameter regions. Hence, it is necessary to understand how, where, and what types of such chaotic dynamics occur before they can be used in practical devices. We consider a pair of coupled differential equations that describes the interactions of two optical beams in a single-mode resonator with symmetric pumping. Recently, it was shown that this system exhibits a wide range of fascinating behaviors, including bistability, symmetry breaking, chaos, and self-switching oscillations. We employ here a dynamical system approach to identify, delimit, and explain the regions where such different behaviors can be observed. Specifically, we find that different kinds of self-switching oscillations are created via the collision of a pair of asymmetric periodic orbits or chaotic attractors at Shilnikov homoclinic bifurcations, which acts as a gluing bifurcation. We present a bifurcation diagram that shows how these global bifurcations are organized by a Belyakov transition point (where the stability of the homoclinic orbit changes). In this way, we map out distinct transitions to different chaotic switching behavior that should be expected from this optical device., Comment: 22 pages, 10 figures

Studying extreme ideas in routine choices and discussions is of utmost importance to understand the increasing polarization in society. In this study, we focus on understanding the generation and influence of extreme ideas in routine conversations which we label "eccentric" ideas. The eccentricity of any idea is defined as the deviation of that idea from the norm of the social neighborhood. We collected and analyzed data from two completely different sources: public social media and online experiments in a controlled environment. We compared the popularity of ideas against their eccentricity to understand individuals' fascination towards eccentricity. We found that more eccentric ideas have a higher probability of getting a greater number of "likes". Additionally, we demonstrate that the social neighborhood of an individual conceals eccentricity changes in one's own opinions and facilitates generation of eccentric ideas at a collective level., Comment: 15 pages, 3 figures

Structure reflects a variety of practices teachers use with the intent to guide students' behavior and increase academic success. A research synthesis was conducted on the role of classroom structure in the academic engagement, disengagement, competence beliefs, and achievement of preschool through high school students. A meta-analysis of 191 samples from 165 correlational studies revealed statistically significant correlations with achievement (0.11), engagement (0.28), and competence beliefs (0.22), and a statistically non-significant relationship with disengagement (-0.08). A meta-analysis of 71 samples from 46 structure intervention studies revealed a positive statistically significant average effect (g) on achievement (0.33), engagement (0.46), and disengagement (-0.34), but a statistically non-significant effect for competence beliefs (0.26). Consistent with a dual process model of engagement, associations were stronger for engagement than disengagement. Results related to variation suggested some universality, particularly across grade levels, and underscored the importance of emphasizing anticipatory strategies, minimizing the controlling aspects of structure, and considering the broader context, including the country context, income background of students, or whether structure is paired with other psychological supports. Methodological features also explained variation, highlighting the importance of using methods that center teachers' and students' experiences and align with the nature of the focal outcome. [This paper will be published in "Educational Psychologist."]

We study different types of solitons of a generalized nonlinear Schr\"odinger equation (GNLSE) that models optical pulses traveling down an optical waveguide with quadratic as well as quartic dispersion. A traveling-wave ansatz transforms this partial differential equation into a fourth-order nonlinear ordinary differential equation (ODE) that is Hamiltonian and has two reversible symmetries. Homoclinic orbits of the ODE that connect the origin to itself represent solitons of the GNLSE, and this allows us to study the existence and organization of solitons with advanced numerical tools for the detection and continuation of connecting orbits. In this way, we establish the existence of connections from one periodic orbit to another, called PtoP connections. They give rise to families of homoclinic orbits to either of the two periodic orbits; in the GNLSE they correspond to generalized solitons with oscillating tails whose amplitude does not decay but reaches a nonzero limit. Moreover, PtoP connections can be found in the energy level of the origin, where connections between this equilibrium and a given periodic orbit, called EtoP connections, are known to organize families of solitons. As we show here, EtoP and PtoP cycles can be assembled into different types of heteroclinic cycles that give rise to additional families of homoclinic orbits to the origin. In the GNLSE, these correspond to multi-oscillation solitons that feature several episodes of different oscillations in between their decaying tails. As for solitons organized by EtoP connections only, multi-oscillation solitons are shown to be an integral part of the phenomenon of truncated homoclinic snaking., Comment: 25 Pages, 13 figures

The social brain hypothesis states that the relative size of the neocortex is larger for species with higher social complexity as a result of evolution. Various lines of empirical evidence have supported the social brain hypothesis, including evidence from the structure of social networks. Social complexity may itself positively impact cooperation among individuals, which occurs across different animal taxa and is a key behavior for successful group living. Theoretical research has shown that particular structures of social networks foster cooperation more easily than others. Therefore, we hypothesized that species with a relatively large neocortex tend to form social networks that better enable cooperation. In the present study, we combine data on brain and body mass, data on social networks, and theory on the evolution of cooperation on networks to test this hypothesis in primates. We have found a positive effect of brain size on cooperation in social networks even after controlling for the effect of other structural properties of networks that are known to promote cooperation., Comment: 14 pages, 4 figures

Randomized smoothing is a leading approach for constructing classifiers that are certifiably robust against adversarial examples. Existing work on randomized smoothing has focused on classifiers with continuous inputs, such as images, where $\ell_p$-norm bounded adversaries are commonly studied. However, there has been limited work for classifiers with discrete or variable-size inputs, such as for source code, which require different threat models and smoothing mechanisms. In this work, we adapt randomized smoothing for discrete sequence classifiers to provide certified robustness against edit distance-bounded adversaries. Our proposed smoothing mechanism randomized deletion (RS-Del) applies random deletion edits, which are (perhaps surprisingly) sufficient to confer robustness against adversarial deletion, insertion and substitution edits. Our proof of certification deviates from the established Neyman-Pearson approach, which is intractable in our setting, and is instead organized around longest common subsequences. We present a case study on malware detection--a binary classification problem on byte sequences where classifier evasion is a well-established threat model. When applied to the popular MalConv malware detection model, our smoothing mechanism RS-Del achieves a certified accuracy of 91% at an edit distance radius of 128 bytes., Comment: Final camera-ready version for NeurIPS 2023. 36 pages, 7 figures, 12 tables. Includes 20 pages of appendices. Code available at https://github.com/Dovermore/randomized-deletion

Entity resolution (record linkage or deduplication) is the process of identifying and linking duplicate records in databases. In this paper, we propose a Bayesian graphical approach for entity resolution that links records to latent entities, where the prior representation on the linkage structure is exchangeable. First, we adopt a flexible and tractable set of priors for the linkage structure, which corresponds to a special class of random partition models. Second, we propose a more realistic distortion model for categorical/discrete record attributes, which corrects a logical inconsistency with the standard hit-miss model. Third, we incorporate hyperpriors to improve flexibility. Fourth, we employ a partially collapsed Gibbs sampler for inferential speedups. Using a selection of private and nonprivate data sets, we investigate the impact of our modeling contributions and compare our model with two alternative Bayesian models. In addition, we conduct a simulation study for household survey data, where we vary distortion, duplication rates and data set size. We find that our model performs more consistently than the alternatives across a variety of scenarios and typically achieves the highest entity resolution accuracy (F1 score). Open source software is available for our proposed methodology, and we provide a discussion regarding our work and future directions., Comment: 27 pages, 4 figures, 3 tables. Includes 37 pages of appendices. This is an accepted manuscript to be published in the Journal of Survey Statistics and Methodology

Excitability, encountered in numerous fields from biology to neurosciences and optics, is a general phenomenon characterized by an all-or-none response of a system to an external perturbation. When subject to delayed feedback, excitable systems can sustain multistable pulsing regimes, which are either regular or irregular time sequences of pulses reappearing every delay time. Here, we investigate an excitable microlaser subject to delayed optical feedback and study the emergence of complex pulsing dynamics, including periodic, quasiperiodic and irregular pulsing regimes. This work is motivated by experimental observations showing these different types of pulsing dynamics. A suitable mathematical model, written as a system of delay differential equations, is investigated through an in-depth bifurcation analysis. We demonstrate that resonance tongues play a key role in the emergence of complex dynamics, including non-equidistant periodic pulsing solutions and chaotic pulsing. The structure of resonance tongues is shown to depend very sensitively on the pump parameter. Successive saddle transitions of bounding saddle-node bifurcations constitute a merging process that results in unexpectedly large locking regions, which subsequently disconnect from the relevant torus bifurcation curve; the existence of such unconnected regions of periodic pulsing is in excellent agreement with experimental observations. As we show, the transition to unconnected resonance regions is due to a general mechanism: the interaction of resonance tongues locally at an extremum of the rotation number on a torus bifurcation curve. We present and illustrate the two generic cases of disconnecting and of disappearing resonance tongues. Moreover, we show how a maximum and a minimum of the rotation number appears naturally when two torus bifurcation curves undergo a saddle transition (where they connect differently)., Comment: 11 figures

Successfully anticipating sudden major changes in complex systems is a practical concern. Such complex systems often form a heterogeneous network, which may show multistage transitions in which some nodes experience a regime shift earlier than others as an environment gradually changes. Here we investigate early warning signals for networked systems undergoing a multistage transition. We found that knowledge of both the ongoing multistage transition and network structure enables us to calculate effective early warning signals for multistage transitions. Furthermore, we found that small subsets of nodes could anticipate transitions as well as or even better than using all the nodes. Even if we fix the network and dynamical system, no single best subset of nodes provides good early warning signals, and a good choice of sentinel nodes depends on the tipping direction and the current stage of the dynamics within a multistage transition, which we systematically characterize., Comment: Updated text and figures with additional analysis; 45 pages w/ main text and SI in one PDF

Many complex dynamical systems in the real world, including ecological, climate, financial, and power-grid systems, often show critical transitions, or tipping points, in which the system's dynamics suddenly transit into a qualitatively different state. In mathematical models, tipping points happen as a control parameter gradually changes and crosses a certain threshold. Tipping elements in such systems may interact with each other as a network, and understanding the behavior of interacting tipping elements is a challenge because of the high dimensionality originating from the network. Here we develop a degree-based mean-field theory for a prototypical double-well system coupled on a network with the aim of understanding coupled tipping dynamics with a low-dimensional description. The method approximates both the onset of the tipping point and the position of equilibria with a reasonable accuracy. Based on the developed theory and numerical simulations, we also provide evidence for multistage tipping point transitions in networks of double-well systems., Comment: 5 figures

Collective idea generation and innovation processes are complex and dynamic, involving a large amount of qualitative narrative information that is difficult to monitor, analyze, and visualize using traditional methods. In this study, we developed three new visualization methods for collective idea generation and innovation processes and applied them to data from online social network experiments. The first visualization is the Idea Cloud, which helps monitor collective idea posting activity and intuitively tracks idea clustering and transition. The second visualization is the Idea Geography, which helps understand how the idea space and its utility landscape are structured and how collaboration was performed in that space. The third visualization is the Idea Network, which connects idea dynamics with the social structure of the people who generated them, displaying how social influence among neighbors may have affected collaborative activities and where innovative ideas arose and spread in the social network.

We consider the open two-site Bose-Hubbard dimer, a well-known quantum mechanical model that has been realised recently for photons in two coupled photonic crystal nanocavities. The system is described by a Lindblad master equation which, for large numbers of photons, gives rise to a limiting semiclassical model in the form of a four-dimensional vector field. From the situation where both sites trap the same amount of photons under symmetric pumping, one encounters a transition that involves symmetry breaking, the creation of periodic oscillations and multistability as the pump strength is increased. We show that the associated one-parameter bifurcation diagram of the semiclassical model captures the essence of statistical properties of computed quantum trajectories as the pump strength is increased. Even for small numbers of photons, the fingerprint of the semiclassical bifurcations can be recognised reliably in observables of quantum trajectories., Comment: 19 pages, 10 figures

The right to erasure requires removal of a user's information from data held by organizations, with rigorous interpretations extending to downstream products such as learned models. Retraining from scratch with the particular user's data omitted fully removes its influence on the resulting model, but comes with a high computational cost. Machine "unlearning" mitigates the cost incurred by full retraining: instead, models are updated incrementally, possibly only requiring retraining when approximation errors accumulate. Rapid progress has been made towards privacy guarantees on the indistinguishability of unlearned and retrained models, but current formalisms do not place practical bounds on computation. In this paper we demonstrate how an attacker can exploit this oversight, highlighting a novel attack surface introduced by machine unlearning. We consider an attacker aiming to increase the computational cost of data removal. We derive and empirically investigate a poisoning attack on certified machine unlearning where strategically designed training data triggers complete retraining when removed., Comment: Align with camera-ready submission to AAAI-22. Changes include: switched to row-wise normalization in Algorithm 3, added link to GitHub repository, added Appendix C with additional results on long-term effectiveness

The Bose--Hubbard dimer model is a celebrated fundamental quantum mechanical model that accounts for the dynamics of bosons at two interacting sites. It has been realized experimentally by two coupled, driven and lossy photonic crystal nanocavities, which are optical devices that operate with only a few hundred photons due to their extremely small size. Our work focuses on characterizing the different dynamics that arise in the semiclassical approximation of such driven-dissipative photonic Bose--Hubbard dimers. Mathematically, this system is a four-dimensional autonomous vector field that describes two specific coupled oscillators, where both the amplitude and the phase are important. We perform a bifurcation analysis of this system to identify regions of different behavior as the pump power $f$ and the detuning $\delta$ of the driving signal are varied, for the case of fixed positive coupling. The bifurcation diagram in the $(f,\delta)$-plane is organized by two points of codimension-two bifurcations -- a $Z_2$-equivariant homoclinic flip bifurcation and a Bykov T-point -- and provides a roadmap for the observable dynamics, including different types of chaotic behavior. To illustrate the overall structure and different accumulation processes of bifurcation curves and associated regions, our bifurcation analysis is complemented by the computation of kneading invariants and of maximum Lyapunov exponents in the $(f,\delta)$-plane. The bifurcation diagram displays a menagerie of dynamical behavior and offers insights into the theory of global bifurcations in a four-dimensional phase space, including novel bifurcation phenomena such as degenerate singular heteroclinic cycles., Comment: 44 pages, 26 figures/