Search

We propose an algorithm for counting the number of cycles under local differential privacy for degeneracy-bounded input graphs. Numerous studies have focused on counting the number of triangles under the privacy notion, demonstrating that the expected $\ell_2$-error of these algorithms is $\Omega(n^{1.5})$, where $n$ is the number of nodes in the graph. When parameterized by the number of cycles of length four ($C_4$), the best existing triangle counting algorithm has an error of $O(n^{1.5} + \sqrt{C_4}) = O(n^2)$. In this paper, we introduce an algorithm with an expected $\ell_2$-error of $O(\delta^{1.5} n^{0.5} + \delta^{0.5} d_{\max}^{0.5} n^{0.5})$, where $\delta$ is the degeneracy and $d_{\max}$ is the maximum degree of the graph. For degeneracy-bounded graphs ($\delta \in \Theta(1)$) commonly found in practical social networks, our algorithm achieves an expected $\ell_2$-error of $O(d_{\max}^{0.5} n^{0.5}) = O(n)$. Our algorithm's core idea is a precise count of triangles following a preprocessing step that approximately sorts the degree of all nodes. This approach can be extended to approximate the number of cycles of length $k$, maintaining a similar $\ell_2$-error, namely $O(\delta^{(k-2)/2} d_{\max}^{0.5} n^{(k-2)/2} + \delta^{k/2} n^{(k-2)/2})$ or $O(d_{\max}^{0.5} n^{(k-2)/2}) = O(n^{(k-1)/2})$ for degeneracy-bounded graphs.

We present a thorough analysis of computing the Generalized Alignment Index (GALI), a rapid and effective chaos indicator, through a simple multi-particle approach, which avoids the use of variational equations. We develop a theoretical leading-order estimation of the error in the computed GALI for both the variational method (VM) and the multi-particle method (MPM), and confirm its predictions through extensive numerical simulations of two well-known Hamiltonian models: the H\'enon-Heiles and the $\beta$-Fermi-Pasta-Ulam-Tsingou systems. For these models the GALIs of several orders are computed and the MPM results are compared to the VM outcomes. The dependence of the accuracy of the MPM on the renormalization time, integration time step, as well as the deviation vector size, is studied in detail. We find that the implementation if the MPM in double machine precision ($\varepsilon \approx 10^{-16}$) is reliable for deviation vector magnitudes centred around $d_0\approx \varepsilon^{1/2}$, renormalization times $\tau \lesssim 1$, and relative energy errors $E_r \lesssim \varepsilon^{1/2}$. These results are valid for systems with many degrees of freedom and for several orders of the GALIs, with the MPM particularly capturing very accurately the $\textrm{GALI}_2$ behavior. Our results show that the computation of the GALIs by the MPM is a robust and efficient method for investigating the global chaotic dynamics of autonomous Hamiltonian systems, something which is of distinct importance in cases where it is difficult to explicitly write the system's variational equation or when these equations are too cumbersome., Comment: 17 pages, 13 figures

We propose a new statistical reduced complexity climate model. The centerpiece of the model consists of a set of physical equations for the global climate system which we show how to cast in non-linear state space form. The parameters in the model are estimated using the method of maximum likelihood with the likelihood function being evaluated by the extended Kalman filter. Our statistical framework is based on well-established methodology and is computationally feasible. In an empirical analysis, we estimate the parameters for a data set comprising the period 1959-2022. A likelihood ratio test sheds light on the most appropriate equation for converting the level of atmospheric concentration of carbon dioxide into radiative forcing. Using the estimated model, and different future paths of greenhouse gas emissions, we project global mean surface temperature until the year 2100. Our results illustrate the potential of combining statistical modelling with physical insights to arrive at rigorous statistical analyses of the climate system.

Projecting sea-level change in various climate-change scenarios typically involves running forward simulations of the Earth's gravitational, rotational and deformational (GRD) response to ice mass change, which requires high computational cost and time. Here we build neural-network emulators of sea-level change at 27 coastal locations, due to the GRD effects associated with future Antarctic Ice Sheet mass change over the 21st century. The emulators are based on datasets produced using a numerical solver for the static sea-level equation and published ISMIP6-2100 ice-sheet model simulations referenced in the IPCC AR6 report. We show that the neural-network emulators have an accuracy that is competitive with baseline machine learning emulators. In order to quantify uncertainty, we derive well-calibrated prediction intervals for simulated sea-level change via a linear regression postprocessing technique that uses (nonlinear) machine learning model outputs, a technique that has previously been applied to numerical climate models. We also demonstrate substantial gains in computational efficiency: a feedforward neural-network emulator exhibits on the order of 100 times speedup in comparison to the numerical sea-level equation solver that is used for training.

We introduce "pointer-guided segment ordering" (SO), a novel pre-training technique aimed at enhancing the contextual understanding of paragraph-level text representations in large language models. Our methodology leverages a self-attention-driven pointer network to restore the original sequence of shuffled text segments, addressing the challenge of capturing the structural coherence and contextual dependencies within documents. This pre-training approach is complemented by a fine-tuning methodology that incorporates dynamic sampling, augmenting the diversity of training instances and improving sample efficiency for various downstream applications. We evaluate our method on a diverse set of datasets, demonstrating its efficacy in tasks requiring sequential text classification across scientific literature and financial reporting domains. Our experiments show that pointer-guided pre-training significantly enhances the model's ability to understand complex document structures, leading to state-of-the-art performance in downstream classification tasks., Comment: 17 pages, 3 figures, 5 tables, accepted at ECML-PKDD 2024

This study presents a statistical time-domain approach for identifying transitions between climate states, referred to as breakpoints, using well-established econometric tools. We analyze a 67.1 million year record of the oxygen isotope ratio delta-O-18 derived from benthic foraminifera. The dataset is presented in Westerhold et al. (2020), where the authors use recurrence analysis to identify six climate states. Fixing the number of breakpoints to five, our procedure results in breakpoint estimates that closely align with those identified by Westerhold et al. (2020). By treating the number of breakpoints as a parameter to be estimated, we provide the statistical justification for more than five breakpoints in the time series. Further, our approach offers the advantage of constructing confidence intervals for the breakpoints, and it allows for testing the number of breakpoints present in the time series., Comment: Authors in alphabetical order

Time series analysis of delta-O-18 and delta-C-13 measurements from benthic foraminifera for purposes of paleoclimatology is challenging. The time series reach back tens of millions of years, they are relatively sparse in the early record and relatively dense in the later, the time stamps of the observations are not evenly spaced, and there are instances of multiple different observations at the same time stamp. The time series appear non-stationary over most of the historical record with clearly visible temporary trends of varying directions. In this paper, we propose a continuous-time state-space framework to analyze the time series. State space models are uniquely suited for this purpose, since they can accommodate all the challenging features mentioned above. We specify univariate models and joint bivariate models for the two time series of delta-O-18 and delta-C-13. The models are estimated using maximum likelihood by way of the Kalman filter recursions. The suite of models we consider has an interpretation as an application of the Butterworth filter. We propose model specifications that take the origin of the data from different studies into account and that allow for a partition of the total period into sub-periods reflecting different climate states. The models can be used, for example, to impute evenly time-stamped values by way of Kalman filtering. They can also be used, in future work, to analyze the relation to proxies for CO2 concentrations., Comment: Authors in alphabetical order

We investigate how the phase space structure of a 3D autonomous Hamiltonian system evolves across a series of successive 2D and 3D pitchfork and period-doubling bifurcations, as the transition of the parent families of periodic orbits (POs) from stability to simple instability leads to the creation of new stable POs. Our research illustrates the consecutive alterations in the phase space structure near POs as the stability of the main family of POs changes. This process gives rise to new families of POs within the system, either maintaining the same or exhibiting higher multiplicity compared to their parent families. Tracking such a phase space transformation is challenging in a 3D system. By utilizing the color and rotation technique to visualize the 4D Poincar'e surfaces of section of the system, i.e. projecting them onto a 3D subspace and employing color to represent the fourth dimension, we can identify distinct structural patterns. Perturbations of parent and bifurcating stable POs result in the creation of tori characterized by a smooth color variation on their surface. Furthermore, perturbations of simple unstable parent POs beyond the bifurcation point which lead to the birth of new stable families of POs, result in the formation of figure-8 structures of smooth color variations. These figure-8 formations surround well-shaped tori around the bifurcated stable POs, losing their well-defined forms for energies further away from the bifurcation point. We also observe that even slight perturbations of highly unstable POs create a cloud of mixed color points, which rapidly move away from the location of the PO. Our study introduces, for the first time, a systematic visualization of 4D surfaces of section within the vicinity of higher multiplicity POs. It elucidates how, in these cases, the coexistence of regular and chaotic orbits contributes to shaping the phase space landscape., Comment: 24 pages, 16 figures, accepted for publication in International Journal of Bifurcation and Chaos

We suggest the use of hash functions to cut down the communication costs when counting subgraphs under edge local differential privacy. While various algorithms exist for computing graph statistics, including the count of subgraphs, under the edge local differential privacy, many suffer with high communication costs, making them less efficient for large graphs. Though data compression is a typical approach in differential privacy, its application in local differential privacy requires a form of compression that every node can reproduce. In our study, we introduce linear congruence hashing. With a sampling rate of $s$, our method can cut communication costs by a factor of $s^2$, albeit at the cost of increasing variance in the published graph statistic by a factor of $s$. The experimental results indicate that, when matched for communication costs, our method achieves a reduction in the $\ell_2$-error for triangle counts by up to 1000 times compared to the performance of leading algorithms., Comment: 13 pages, 3 figures

We propose a non-linear state-space model to examine the relationship between CO$_2$ emissions, energy sources, and macroeconomic activity, using data from 1971 to 2019. CO$_2$ emissions are modeled as a weighted sum of fossil fuel use, with emission conversion factors that evolve over time to reflect technological changes. GDP is expressed as the outcome of linearly increasing energy efficiency and total energy consumption. The model is estimated using CO$_2$ data from the Global Carbon Budget, GDP statistics from the World Bank, and energy data from the International Energy Agency (IEA). Projections for CO$_2$ emissions and GDP from 2020 to 2100 from the model are based on energy scenarios from the Shared Socioeconomic Pathways (SSP) and the IEA's Net Zero roadmap. Emissions projections from the model are consistent with these scenarios but predict lower GDP growth. An alternative model version, assuming exponential energy efficiency improvement, produces GDP growth rates more in line with the benchmark projections. Our results imply that if internationally agreed net-zero objectives are to be fulfilled and economic growth is to follow SSP or IEA scenarios, then drastic changes in energy efficiency, not consistent with historical trends, are needed.

The global fraction of anthropogenically emitted carbon dioxide (CO$_2$) that stays in the atmosphere, the CO$_2$ airborne fraction, has been fluctuating around a constant value over the period 1959 to 2022. The consensus estimate of the airborne fraction is around $44\%$; the remaining $56\%$ is absorbed by the oceanic and terrestrials biospheres. In this study, we show that the conventional estimator of the airborne fraction, based on a ratio of changes in atmospheric CO$_2$ concentrations and CO$_2$ emissions, suffers from a number of statistical deficiencies, such as non-existence of moments and a non-Gaussian limiting distribution. We propose an alternative regression-based estimator of the airborne fraction that does not suffer from these deficiencies. We show that the regression-based estimator has a Gaussian limiting distribution and reduces estimation uncertainty substantially. Our empirical analysis leads to an estimate of the airborne fraction over 1959--2022 of $47.0\%$ ($\pm 1.1\%$; $1 \sigma$), implying a higher, and better constrained, estimate than the current consensus. Using climate model output, we show that a regression-based approach provides sensible estimates of the airborne fraction, also in future scenarios where emissions are at or near zero.

In this paper, we give an algorithm to publish the number of paths and Katz centrality under the local differential privacy (LDP), providing a thorough theoretical analysis. Although various works have already introduced subgraph counting algorithms under LDP, they have primarily concentrated on subgraphs of up to five nodes. The challenge in extending this to larger subgraphs is the cumulative and exponential growth of noise as the subgraph size increases in any publication under LDP. We address this issue by proposing an algorithm to publish the number of paths that start at every node in the graph, leading to an algorithm that publishes the Katz centrality of all nodes. This algorithm employs multiple rounds of communication and the clipping technique. Both our theoretical and experimental assessments indicate that our algorithm exhibits acceptable bias and variance, considerably less than an algorithm that bypasses clipping. Furthermore, our Katz centrality estimation is able to recall up to 90% of the nodes with the highest Katz centrality.

Ever-larger language models with ever-increasing capabilities are by now well-established text processing tools. Alas, information extraction tasks such as named entity recognition are still largely unaffected by this progress as they are primarily based on the previous generation of encoder-only transformer models. Here, we propose a simple yet effective approach, Informed Named Entity Recognition Decoding (iNERD), which treats named entity recognition as a generative process. It leverages the language understanding capabilities of recent generative models in a future-proof manner and employs an informed decoding scheme incorporating the restricted nature of information extraction into open-ended text generation, improving performance and eliminating any risk of hallucinations. We coarse-tune our model on a merged named entity corpus to strengthen its performance, evaluate five generative language models on eight named entity recognition datasets, and achieve remarkable results, especially in an environment with an unknown entity class set, demonstrating the adaptability of the approach., Comment: 12 pages, 2 figures, 4 tables

Auditing financial documents is a very tedious and time-consuming process. As of today, it can already be simplified by employing AI-based solutions to recommend relevant text passages from a report for each legal requirement of rigorous accounting standards. However, these methods need to be fine-tuned regularly, and they require abundant annotated data, which is often lacking in industrial environments. Hence, we present ZeroShotALI, a novel recommender system that leverages a state-of-the-art large language model (LLM) in conjunction with a domain-specifically optimized transformer-based text-matching solution. We find that a two-step approach of first retrieving a number of best matching document sections per legal requirement with a custom BERT-based model and second filtering these selections using an LLM yields significant performance improvements over existing approaches., Comment: Accepted at DocEng 2023, 4 pages, 1 figure, 2 tables

We present sustainAI, an intelligent, context-aware recommender system that assists auditors and financial investors as well as the general public to efficiently analyze companies' sustainability reports. The tool leverages an end-to-end trainable architecture that couples a BERT-based encoding module with a multi-label classification head to match relevant text passages from sustainability reports to their respective law regulations from the Global Reporting Initiative (GRI) standards. We evaluate our model on two novel German sustainability reporting data sets and consistently achieve a significantly higher recommendation performance compared to multiple strong baselines. Furthermore, sustainAI is publicly available for everyone at https://sustain.ki.nrw/., Comment: Accepted at ICAIL 2023, 5 pages, 3 figure, 3 tables

We investigate the ability of simple diagnostics based on Lagrangian descriptor (LD) computations of initially nearby orbits to detect chaos in conservative dynamical systems with phase space dimensionality higher than two. In particular, we consider the recently introduced methods of the difference ($D_L^n$) and the ratio ($R_L^n$) of the LDs of neighboring orbits, as well as a quantity ($S_L^n$) related to the finite-difference second spatial derivative of the LDs, and use them to determine the chaotic or regular nature of ensembles of orbits of a prototypical area-preserving map model, the 4-dimensional (4D) symplectic standard map. Using the distributions of the indices' values we determine appropriate thresholds to discriminate between regular and chaotic orbits, and compare the obtained characterization against that achieved by the Smaller Alignment Index (SALI) method of chaos detection, by recording the percentage agreement $P_A$ between the two classifications. We study the influence of various factors on the performance of these indices, and show that the increase of the final number of orbit iterations T and the order n of the indices (i.e. the dimensionality of the space where the considered nearby orbits lie), as well as the decrease of the distance $\sigma$ of neighboring orbits, increase the $P_A$ values along with the required computational effort. Balancing between these two factors we find appropriate T, n and $\sigma$ values, which allow the efficient use of the $D_L^n$, $R_L^n$ and $S_L^n$ indices as short time and computationally cheap chaos diagnostics achieving $P_A \gtrsim 90 \%$, with $D_L^n$ and $S_L^n$ having larger $P_A$ values than $R_L^n$. Our results show that the three LDs-based indices perform better for systems with large percentages of chaotic orbits.

We introduce and demonstrate the usage of the origin-fate map (OFM) as a tool for the detailed investigation of phase space transport in reactant-product type systems. For these systems, which exhibit clearly defined start and end states, it is possible to build a comprehensive picture of the lobe dynamics by considering backward and forward integration of sets of initial conditions to index their origin and fate. We illustrate the method and its utility in the study of a two degrees of freedom caldera potential with four exits, demonstrating that the OFM not only recapitulates results from classical manifold theory, but even provides more detailed information about complex lobe structures. The OFM allows the detection of dynamically significant transitions caused by the creation of new lobes, and is also able to guide the prediction of the position of unstable periodic orbits (UPOs). Further, we compute the OFM on the periodic orbit dividing surface (PODS) associated with the transition state of a caldera entrance, which allows for a powerful analysis of reactive trajectories. The intersection of the manifolds corresponding to this UPO with other manifolds in the phase space results in the appearance of lobes on the PODS, which are directly classified by the OFM. This allows computations of branching ratios and the exploration of a fractal cascade of lobes as the caldera is stretched, which results in fluctuations in the branching ratio and chaotic selectivity. The OFM is found to be a simple and very useful tool with a vast range of descriptive and quantitative applications., Comment: 11 pages, 10 figures

We introduce KPI-Check, a novel system that automatically identifies and cross-checks semantically equivalent key performance indicators (KPIs), e.g. "revenue" or "total costs", in real-world German financial reports. It combines a financial named entity and relation extraction module with a BERT-based filtering and text pair classification component to extract KPIs from unstructured sentences before linking them to synonymous occurrences in the balance sheet and profit & loss statement. The tool achieves a high matching performance of $73.00$% micro F$_1$ on a hold out test set and is currently being deployed for a globally operating major auditing firm to assist the auditing procedure of financial statements., Comment: Accepted at BigData 2022, 10 pages, 3 figure, 5 tables

We introduce KPI-EDGAR, a novel dataset for Joint Named Entity Recognition and Relation Extraction building on financial reports uploaded to the Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system, where the main objective is to extract Key Performance Indicators (KPIs) from financial documents and link them to their numerical values and other attributes. We further provide four accompanying baselines for benchmarking potential future research. Additionally, we propose a new way of measuring the success of said extraction process by incorporating a word-level weighting scheme into the conventional F1 score to better model the inherently fuzzy borders of the entity pairs of a relation in this domain., Comment: Accepted at ICMLA 2022, 6 pages, 5 tables

We present and validate simple and efficient methods to estimate the chaoticity of orbits in low dimensional dynamical systems from computations of Lagrangian descriptors (LDs) on short time scales. Two quantities are proposed for determining the chaotic or regular nature of orbits in a system's phase space, which are based on the values of the LDs of these orbits and of nearby ones: The difference (DNLD) and ratio (RNLD) of neighboring orbits' LDs. We find that, typically, these indicators are able to correctly characterize the chaotic or regular nature of orbits to better than 90 % agreement with results obtained by implementing the Smaller Alignment Index (SALI) method, which is a well established chaos detection technique. Further investigating the performance of the two introduced quantities we discuss the effects of the total integration time and of the spacing between the used neighboring orbits on the accuracy of the methods, finding that even typical short time, coarse grid LD computations are sufficient to provide a reliable quantification of a system's chaotic component, using less CPU time than the SALI. In addition to quantifying chaos, the introduced indicators have the ability to reveal details about the systems' local and global chaotic phase space structure. Our findings clearly suggest that LDs can also be used to quantify and investigate chaos in continuous and discrete low dimensional dynamical systems.

Understanding the inherent timescales of large bubbles in DNA is critical to a thorough comprehension of its physicochemical characteristics, as well as their potential role on helix opening and biological function. In this work we employ the coarse-grained Peyrard-Bishop-Dauxois model of DNA to study relaxation dynamics of large bubbles in homopolymer DNA, using simulations up to the microsecond time scale. By studying energy autocorrelation functions of relatively large bubbles inserted into thermalised DNA molecules, we extract characteristic relaxation times from the equilibration process for both adenine-thymine (AT) and guanine-cytosine (GC) homopolymers. Bubbles of different amplitudes and widths are investigated through extensive statistics and appropriate fittings of their relaxation. Characteristic relaxation times increase with bubble height and width. We show that, within the model, relaxation times are two orders of magnitude longer in GC sequences than in AT sequences. Overall, our results confirm that large bubbles leave a lasting impact on the molecule's dynamics, for times between 0.4-200ns depending on the homopolymer type and bubble shape, thus clearly affecting long-time evolutions of the molecule., Comment: 11 pages, 6 figures

We present KPI-BERT, a system which employs novel methods of named entity recognition (NER) and relation extraction (RE) to extract and link key performance indicators (KPIs), e.g. "revenue" or "interest expenses", of companies from real-world German financial documents. Specifically, we introduce an end-to-end trainable architecture that is based on Bidirectional Encoder Representations from Transformers (BERT) combining a recurrent neural network (RNN) with conditional label masking to sequentially tag entities before it classifies their relations. Our model also introduces a learnable RNN-based pooling mechanism and incorporates domain expert knowledge by explicitly filtering impossible relations. We achieve a substantially higher prediction performance on a new practical dataset of German financial reports, outperforming several strong baselines including a competing state-of-the-art span-based entity tagging approach., Comment: Accepted at ICPR 2022, 8 pages, 1 figure, 6 tables