arXiv subjects

Karoline Wiesner

Publications and source records attributed to Karoline Wiesner.

At least 19 recordsLinked to original sources

arXiv2026

A Generalized Gompertz Law for the Rise and Fall of Empires

Human societies tend to evolve toward increasing scale and complexity, often followed by stagnation and decline. Historical empires provide a prominent example of this trajectory; however, existing quantitative models capture expansion and rarely account for decline within a unified framework. Here, we introduce a minimal dynamical model that describes both growth and collapse in a single formalism. The model yields a closed-form trajectory that generalizes the Gompertz law and admits equivalent maximum-entropy and optimal-control formulations. We show that this functional form captures the rise and fall of fourteen diverse empires across widely varying time scales. Despite substantial historical heterogeneity, all cases follow a common asymmetric rise-peak-decline pattern governed by a small number of interpretable parameters. We identify the mechanism driving this behavior as the chronophage: an exponentially accumulating burden of administrative and coordination costs that offsets the gains of expansion. These findings uncover a general quantitative principle, yielding new insights into how increasing complexity constrains large-scale social systems.

physics.soc-ph↗

arXiv2026

Cascading Through the Hierarchy: Regularizer-Induced Feature Detection as Phase Transitions in Deep Linear Neural Networks

A scientific theory of deep learning, comprising learning dynamics and statistical properties of learned models, is rapidly gaining attention. One of the corner stones of this development are analytically solvable toy models, allowing for the fully tractable analysis of the learning dynamics. Here we analytically investigate such a toy model using the regularization strength as a tunable external parameter - akin to external fields in statistical physics. In previous studies, (i) an onset of learning transition was predicted analytically and (ii) it was phenomenologically/numerically established that tuning the regularization strength can result in a cascade of phase transitions. The number of those transitions was linked to the geometry of the loss landscape determined by the model complexity. Setting up a rigorous framework underpinning the previous numerical observations, our investigation reveals a precise connection between those cascades of phase transitions, learnable features and the underlying geometry. We provide analytic predictions of these phase transitions as well as tractable order parameters related to learned features. At the level of the minimal model, we connect this macroscopic perspective (that can be condensed into an effective description) to the microscopic perspective in terms of the geometry of the loss landscape characterized by the Hessian spectrum. Thus, the presented model provides a platform to explore and sharpen advances made in the scientific theory of deep learning rooted in statistical physics concepts.

cond-mat.stat-mech↗

arXiv2026

Noise-Driven Escape from Metastable Phases explains Grokking in Deep Neural Networks

Deep neural networks (DNNs) exhibit first order phase transitions under variations of the L2 regularization strength, with each transition marking the onset of a new learnable feature. Below a critical regularization strength, all features are in principle learnable, but coexisting metastable states, separated by energy barriers, can trap the network and impede convergence. A strength of DNNs is their ability to generalize. But many open questions remain, among them the origin of so called grokking: the abrupt, delayed onset of generalization after prolonged apparent overfitting. We show for linear DNNs that grokking is consistent with hysteresis in first-order L2 phase transitions: using L2 regularization to engineer deliberate trapping, we demonstrate that a model in a low-accuracy metastable state escapes only when SGD noise drives it across an energy barrier, with escape times following Arrhenius scaling. We reproduce grokking-like delayed convergence across two orders of magnitude in escape time by deliberately trapping models in metastable phases. Using sparse sub-sampling we also reproduce the canonical grokking curve where test error eventually approaches the final training error. Our work suggests that the number of metastable states equals the number of learnable features -- one per singular value of the data covariance -- the potential for hysteresis grows naturally with task complexity. We provide evidence that the same mechanism likely operates in general nonlinear DNNs. Our results provide routes toward more efficient learning schemes.

cs.LG↗

arXiv2026

Critical dynamics govern the evolution of political regimes

The emergence and decline of democratic systems worldwide raises fundamental questions about the dynamics of political change. Contrary to the idea of a stable endpoint of liberal democracy, recent backsliding towards less democratic regimes highlights the non-stationary nature of regime evolution. Here, we analyse the historical trajectories of countries within a two-dimensional regime space derived from the principal components of the Varieties of Democracy dataset. We observe weakly non-ergodic dynamics unfolding in an effective landscape characterised by sparse and shifting basins of stability. Step sizes and sojourn times characterising this dynamics follow heavy-tailed distributions near the critical regime, in which mean values appear to diverge. These facts point to the intermittent and heterogeneous nature of the regime change dynamics. A continuous time random walk model reproduces the dynamics of the three most recent decades with remarkable accuracy. Together, these results suggest that some aspects of political regime evolution follow universal stochastic principles, while remaining punctuated by unique historical pathways.

cond-mat.stat-mech↗

arXiv2026

Nonlinear Dimensionality Reduction with Diffusion Maps in Practice

Diffusion Map is a spectral dimensionality reduction technique which is able to uncover nonlinear submanifolds in high-dimensional data. And, it is increasingly applied across a wide range of scientific disciplines, such as biology, engineering, and social sciences. But data preprocessing, parameter settings and component selection have a significant influence on the resulting manifold, something which has not been comprehensively discussed in the literature so far. We provide a practice oriented review of the Diffusion Map technique, illustrate pitfalls and showcase a recently introduced technique for identifying the most relevant components. Our results show that the first components are not necessarily the most relevant ones.

cs.LG↗

arXiv2025

Phase transitions reveal hierarchical structure in deep neural networks

Training Deep Neural Networks relies on the model converging on a high-dimensional, non-convex loss landscape toward a good minimum. Yet, much of the phenomenology of training remains ill understood. We focus on three seemingly disparate observations: the occurrence of phase transitions reminiscent of statistical physics, the ubiquity of saddle points, and phenomenon of mode connectivity relevant for model merging. We unify these within a single explanatory framework, the geometry of the loss and error landscapes. We analytically show that phase transitions in DNN learning are governed by saddle points in the loss landscape. Building on this insight, we introduce a simple, fast, and easy to implement algorithm that uses the L2 regularizer as a tool to probe the geometry of error landscapes. We apply it to confirm mode connectivity in DNNs trained on the MNIST dataset by efficiently finding paths that connect global minima. We then show numerically that saddle points induce transitions between models that encode distinct digit classes. Our work establishes the geometric origin of key training phenomena in DNNs and reveals a hierarchy of accuracy basins analogous to phases in statistical physics.

cs.LG↗

arXiv2025

Space, time and altruism in pandemics and the climate emergency

Climate change is a global emergency, as was the COVID-19 pandemic. Why was our collective response to COVID-19 so much stronger than our response to the climate emergency, to date? We hypothesize that the answer has to do with the scale of the systems, and not just spatial and temporal scales but also the `altruistic scale' that measures whether an action must rely upon altruistic motives for it to be adopted. We treat COVID-19 and climate change as common pool resource problems that exemplify coupled human-environment systems. We introduce a framework that captures regimes of containment, mitigation, and failure to control. As parameters governing these three scales are varied, it is possible to shift from a COVID-like system to a climate-like system. The framework replicates both inaction in the case of climate change mitigation, as well as the faster response that we exhibited to COVID-19. Our cross-system comparison also suggests actionable ways that cooperation can be improved in large-scale common pool resources problems, like climate change. More broadly, we argue that considering scale and incorporating human-natural system feedbacks are not just interesting special cases within non-cooperative game theory, but rather should be the starting point for the study of altruism and human cooperation.

physics.soc-ph↗

arXiv2025

Phase Transitions between Accuracy Regimes in L2 regularized Deep Neural Networks

Increasing the L2 regularization of Deep Neural Networks (DNNs) causes a first-order phase transition into the under-parametrized phase -- the so-called onset-of learning. We explain this transition via the scalar (Ricci) curvature of the error landscape. We predict new transition points as the data complexity is increased and, in accordance with the theory of phase transitions, the existence of hysteresis effects. We confirm both predictions numerically. Our results provide a natural explanation of the recently discovered phenomenon of '\emph{grokking}' as DNN models getting stuck in a local minimum of the error surface, corresponding to a lower accuracy phase. Our work paves the way for new probing methods of the intrinsic structure of DNNs in and beyond the L2 context.

cs.LG↗

arXiv2024

Unraveling 20th-century political regime dynamics using the physics of diffusion

Uncertainty persists over how and why some countries become democratic and others do not, or why some countries remain democratic and others 'backslide' toward autocracy. Furthermore, while scholars generally agree on the nature of 'democracy' and 'autocracy', the nature of regimes in-between, and changes between them, are much less clear. By applying the spectral dimensionality-reduction technique Diffusion Map to political-science data from the V-Dem project for the period 1900 to 2021, we identify a low-dimensional non-linear manifold on which all electoral regimes move. Using the diffusion equation from statistical physics, we measure the time scale on which countries change their degree of electoral quality, freedom of association, and freedom of expression depending on their position on the manifold. By quantifying the coefficients of the diffusion equation for each country and over time, we show that democracies behave like sub-diffusive (i.e. slow spreading) particles and that autocracies on the verge of collapse behave like super-diffusive (i.e. fast spreading) particles. We show that regimes in-between exhibit diffusion dynamics distinct from autocracies and democracies, and an overall higher instability. Furthermore, we show that a country's position on the manifold and its dynamics are linked to its propensity for civil conflict. Our study pioneers the use of statistical physics in the analysis of political regimes. Our results provide a quantitative foundation for developing theories about what changes during democratization and democratic backsliding, as well as a new framework for regime-transformation and risk-of-conflict assessment.

physics.soc-ph↗

arXiv2024

The principal components of electoral regimes -- Separating autocracies from pseudo-democracies

A critical issue for society today is the emergence and decline of democracy worldwide. It is unclear, however, how democratic features, such as elections and civil liberties, influence this change. Democracy indices, which are the standard tool to study this question, are based on the a priori assumption that improvement in any individual feature strengthens democracy overall. We show that this assumption does not always hold. We use the V-Dem dataset for a quantitative study of electoral regimes worldwide during the 20th century. We find a so-far overlooked trade-off between election quality and civil liberties. In particular, we identify a threshold in the democratisation process at which the correlation between election quality and civil liberties flips from negative to positive. Below this threshold we can thus clearly separate two kinds of non-democratic regimes: autocracies that govern through tightly controlled elections and regimes in which citizens are free but under less certainty -- a distinction that existing democracy indices cannot make. We discuss the stabilising role of election quality uncovered here in the context of the recently observed decline in democracy score of long-standing democracies, so-called `democratic backsliding' or `democratic recession'.

physics.soc-ph↗

arXiv2023

Improving mean-field network percolation models with neighbourhood information

Mean field theory models of percolation on networks provide analytic estimates of network robustness under node or edge removal. We introduce a new mean field theory model based on generating functions that includes information about the tree-likeness of each node's local neighbourhood. We show that our new model outperforms all other generating function models in prediction accuracy when testing their estimates on a wide range of real-world network data. We compare the new model's performance against the recently introduced message passing models and provide evidence that the standard version is also outperformed, while the `loopy' version is only outperformed on a targeted attack strategy. As we show, however, the computational complexity of our model implementation is much lower than that of message passing algorithms. We provide evidence that all discussed models are poor in predicting networks with highly modular structure with dispersed modules, which are also characterised by high mixing times, identifying this as a general limitation of percolation prediction models.

physics.soc-ph↗

arXiv2023

Evaluating The Impact Of Species Specialisation On Ecological Network Robustness Using Analytic Methods

Ecological networks describe the interactions between different species, informing us of how they rely on one another for food, pollination and survival. If a species in an ecosystem is under threat of extinction, it can affect other species in the system and possibly result in their secondary extinction as well. Understanding how (primary) extinctions cause secondary extinctions on ecological networks has been considered previously using computational methods. However, these methods do not provide an explanation for the properties which make ecological networks robust, and can be computationally expensive. We develop a new analytic model for predicting secondary extinctions which requires no non-deterministic computational simulation. Our model can predict secondary extinctions when primary extinctions occur at random or due to some targeting based on the number of links per species or risk of extinction, and can be applied to an ecological network of any number of layers. Using our model, we consider how false positives and negatives in network data affect predictions for network robustness. We have also extended the model to predict scenarios in which secondary extinctions occur once species lose a certain percentage of interaction strength, and to model the loss of interactions as opposed to just species extinction. From our model, it is possible to derive new analytic results such as how ecological networks are most robust when secondary species degree variance is minimised. Additionally, we show that both specialisation and generalisation in distribution of interaction strength can be advantageous for network robustness, depending upon the extinction scenario being considered.

q-bio.PE↗

arXiv2022

Clarifying How Degree Entropies and Degree-Degree Correlations Relate to Network Robustness

It is often claimed that the entropy of a network's degree distribution is a proxy for its robustness. Here, we clarify the link between degree distribution entropy and giant component robustness to node removal by showing that the former merely sets a lower bound to the latter for randomly configured networks when no other network characteristics are specified. Furthermore, we show that, for networks of fixed expected degree that follow degree distributions of the same form, the degree distribution entropy is not indicative of robustness. By contrast, we show that the remaining degree entropy and robustness have a positive monotonic relationship and give an analytic expression for the remaining degree entropy of the log-normal distribution. We also show that degree-degree correlations are not by themselves indicative of a network's robustness for real networks. We propose an adjustment to how mutual information is measured which better encapsulates structural properties related to robustness.

physics.soc-ph↗

arXiv2020

Measuring complexity

Complexity is a multi-faceted phenomenon, involving a variety of features including disorder, nonlinearity, and self-organisation. We use a recently developed rigorous framework for complexity to understand measures of complexity. We illustrate, by example, how features of complexity can be quantified, and we analyse a selection of purported measures of complexity that have found wide application and explain whether and how they measure complexity. We also discuss some of the classic information-theoretic measures from the 1980s and 1990s. This work gives the reader a tool kit for quantifying features of complexity across the sciences.

nlin.AO↗

arXiv2019

Quantum Rate-Distortion Coding of Relevant Information

Rate-distortion theory provides bounds for compressing data produced by an information source to a specified encoding rate that is strictly less than the source's entropy. This necessarily entails some loss, or distortion, between the original source data and the best approximation after decompression. The so-called Information Bottleneck Method is designed to compress only `relevant' information. Which information is relevant is determined by the correlation between the data being compressed and a variable of interest, so-called side information. In this paper, an Information Bottleneck Method is introduced for the compression of quantum data. The channel communication picture is used for compression and decompression. The rate of compression is derived using an entanglement assisted protocol with classical communication, and under an unproved conjecture that the rate function is convex in the distortion parameter. The optimum channel achieving this rate for a given input state is characterised. The conceptual difficulties arising due to differences in the mathematical formalism between quantum and classical probability theory are discussed and solutions are presented.

quant-ph↗

arXiv2019

The thermodynamical cost of some interpretations of quantum theory. Reply to Prunkl and Timpson, and Davidsson

Here we clarify the assumptions made and conclusions reached in our paper "The thermodynamical cost of some interpretations of quantum theory" [Phys. Rev. A 94, 052127 (2016)], at the light of the criticisms of Prunkl and Timpson [Stud. Hist. Philos. Sci. Part B 63, 114 (2018)], and Davidsson (Master thesis, Stockholm University, 2018). We point out some misunderstandings and some weaknesses of the counterexample Prunkl and Timpson present to challenge our conclusion. We thus conclude, once more, that interpretations of quantum theory which consider the probabilities of measurement outcomes to be determined by objective properties of the measured system and satisfy the assumption that the measured system only has finite memory have a thermodynamical cost.

quant-ph↗

arXiv2017

Fluctuations in Single-Shot $ε$-Deterministic Work Extraction

There has been an increasing interest in the quantification of nearly deterministic work extraction from a finite number of copies of microscopic particles in finite time. This paradigm, so called single-shot epsilon-deterministic work extraction, considers processes with small failure probabilities. However, the resulting fluctuations in the extracted work entailed by this failure probability have not been studied before. In the standard thermodynamics paradigm fluctuation theorems are powerful tools to study fluctuating quantities. Given that standard fluctuation theorems are inadequate for a single-shot scenario, here we formulate and prove a fluctuation relation specific to the single-shot epsilon-deterministic work extraction to bridge this gap. Our results are general in the sense that we allow the system to be in contact with the heat bath at all times. As a corollary of our theorem we derive the known bounds on the epsilon-deterministic work.

quant-ph↗

arXiv2016

Thermodynamical cost of some interpretations of quantum theory

The interpretation of quantum theory is one of the longest-standing debates in physics. Type I interpretations see quantum probabilities as determined by intrinsic properties of the observed system. Type II see them as relational experiences between an observer and the system. It is usually believed that a decision between these two options cannot be made simply on purely physical grounds but requires an act of metaphysical judgment. Here we show that, under some assumptions, the problem is decidable using thermodynamics. We prove that type I interpretations are incompatible with the following assumptions: (i) The choice of which measurement is performed can be made randomly and independently of the system under observation, (ii) the system has limited memory, and (iii) Landauer's erasure principle holds.

quant-ph↗