Skip to main content
arXiv is now an independent nonprofit! Learn more

Showing 1–24 of 24 results for author: Cagnetta, F

.
  1. arXiv:2602.07488  [pdf, ps, other

    cs.LG cs.AI stat.ML

    Deriving Neural Scaling Laws from the statistics of natural language

    Authors: Francesco Cagnetta, Allan Raventós, Surya Ganguli, Matthieu Wyart

    Abstract: Despite the fact that experimental neural scaling laws have substantially guided empirical progress in large-scale machine learning, no existing theory can quantitatively predict the exponents of these important laws for any modern LLM trained on any natural language dataset. We provide the first such theory in the case of data-limited scaling laws. We isolate two key statistical properties of lan… ▽ More

    Submitted 2 July, 2026; v1 submitted 7 February, 2026; originally announced February 2026.

    Comments: ICML 2026

  2. arXiv:2602.06065  [pdf, ps, other

    stat.ML cond-mat.dis-nn cs.CL cs.LG

    Deep networks learn to parse uniform-depth context-free languages from local statistics

    Authors: Jack T. Parley, Francesco Cagnetta, Matthieu Wyart

    Abstract: Understanding how the structure of language can be learned from sentences alone is a central question in both cognitive science and machine learning. Studies of the internal representations of Large Language Models (LLMs) support their ability to parse text when predicting the next word, while representing semantic notions independently of surface form. Yet, which data statistics make these feats… ▽ More

    Submitted 1 June, 2026; v1 submitted 31 January, 2026; originally announced February 2026.

    Comments: Accepted as regular paper at ICML 2026

  3. arXiv:2505.07070  [pdf, ps, other

    cs.LG cond-mat.dis-nn stat.ML

    Scaling Laws and Representation Learning in Simple Hierarchical Languages: Transformers vs. Convolutional Architectures

    Authors: Francesco Cagnetta, Alessandro Favero, Antonio Sclocchi, Matthieu Wyart

    Abstract: How do neural language models acquire a language's structure when trained for next-token prediction? We address this question by deriving theoretical scaling laws for neural network performance on synthetic datasets generated by the Random Hierarchy Model (RHM) -- an ensemble of probabilistic context-free grammars designed to capture the hierarchical structure of natural language while remaining a… ▽ More

    Submitted 11 May, 2025; originally announced May 2025.

    Comments: 14 pages, 8 figures

  4. arXiv:2505.07067  [pdf, other

    stat.ML cond-mat.dis-nn cs.LG

    Learning curves theory for hierarchically compositional data with power-law distributed features

    Authors: Francesco Cagnetta, Hyunmo Kang, Matthieu Wyart

    Abstract: Recent theories suggest that Neural Scaling Laws arise whenever the task is linearly decomposed into power-law distributed units. Alternatively, scaling laws also emerge when data exhibit a hierarchically compositional structure, as is thought to occur in language and images. To unify these views, we consider classification and next-token prediction tasks based on probabilistic context-free gramma… ▽ More

    Submitted 11 May, 2025; originally announced May 2025.

  5. arXiv:2502.12089  [pdf, ps, other

    stat.ML cs.LG

    How Compositional Generalization and Creativity Improve as Diffusion Models are Trained

    Authors: Alessandro Favero, Antonio Sclocchi, Francesco Cagnetta, Pascal Frossard, Matthieu Wyart

    Abstract: Natural data is often organized as a hierarchical composition of features. How many samples do generative models need in order to learn the composition rules, so as to produce a combinatorially large number of novel data? What signal in the data is exploited to learn those rules? We investigate these questions in the context of diffusion models both theoretically and empirically. Theoretically, we… ▽ More

    Submitted 4 June, 2025; v1 submitted 17 February, 2025; originally announced February 2025.

    Journal ref: Proceedings of the 42nd International Conference on Machine Learning (ICML), PMLR 267, 2025

  6. arXiv:2406.00048  [pdf, other

    cs.CL cond-mat.dis-nn cs.LG

    Towards a theory of how the structure of language is acquired by deep neural networks

    Authors: Francesco Cagnetta, Matthieu Wyart

    Abstract: How much data is required to learn the structure of a language via next-token prediction? We study this question for synthetic datasets generated via a Probabilistic Context-Free Grammar (PCFG) -- a tree-like generative model that captures many of the hierarchical structures found in natural languages. We determine token-token correlations analytically in our model and show that they can be used t… ▽ More

    Submitted 29 October, 2024; v1 submitted 28 May, 2024; originally announced June 2024.

    Comments: NeurIPS 2024

  7. arXiv:2307.02693  [pdf, other

    cs.LG stat.ML

    Kernels, Data & Physics

    Authors: Francesco Cagnetta, Deborah Oliveira, Mahalakshmi Sabanayagam, Nikolaos Tsilivis, Julia Kempe

    Abstract: Lecture notes from the course given by Professor Julia Kempe at the summer school "Statistical physics of Machine Learning" in Les Houches. The notes discuss the so-called NTK approach to problems in machine learning, which consists of gaining an understanding of generally unsolvable problems by finding a tractable kernel formulation. The notes are mainly focused on practical applications such as… ▽ More

    Submitted 5 July, 2023; originally announced July 2023.

    Comments: These are notes from the lecture of Julia Kempe given at the summer school "Statistical Physics \& Machine Learning", that took place in Les Houches School of Physics in France from 4th to 29th July 2022

  8. arXiv:2307.02129  [pdf, other

    cs.LG cs.CV stat.ML

    How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model

    Authors: Francesco Cagnetta, Leonardo Petrini, Umberto M. Tomasini, Alessandro Favero, Matthieu Wyart

    Abstract: Deep learning algorithms demonstrate a surprising ability to learn high-dimensional tasks from limited examples. This is commonly attributed to the depth of neural networks, enabling them to build a hierarchy of abstract, low-dimensional data representations. However, how many training examples are required to learn such representations remains unknown. To quantitatively study this question, we in… ▽ More

    Submitted 3 July, 2024; v1 submitted 5 July, 2023; originally announced July 2023.

    Comments: 9 pages, 8 figures

    Journal ref: Phys. Rev. X 14, 031001 (2024)

  9. arXiv:2210.01506  [pdf, other

    cs.LG cs.CV

    How deep convolutional neural networks lose spatial information with training

    Authors: Umberto M. Tomasini, Leonardo Petrini, Francesco Cagnetta, Matthieu Wyart

    Abstract: A central question of machine learning is how deep nets manage to learn tasks in high dimensions. An appealing hypothesis is that they achieve this feat by building a representation of the data where information irrelevant to the task is lost. For image datasets, this view is supported by the observation that after (and not before) training, the neural representation becomes less and less sensitiv… ▽ More

    Submitted 23 November, 2022; v1 submitted 4 October, 2022; originally announced October 2022.

  10. arXiv:2208.01003  [pdf, other

    stat.ML cs.LG

    What Can Be Learnt With Wide Convolutional Neural Networks?

    Authors: Francesco Cagnetta, Alessandro Favero, Matthieu Wyart

    Abstract: Understanding how convolutional neural networks (CNNs) can efficiently learn high-dimensional functions remains a fundamental challenge. A popular belief is that these models harness the local and hierarchical structure of natural data such as images. Yet, we lack a quantitative understanding of how such structure affects performance, e.g., the rate of decay of the generalisation error with the nu… ▽ More

    Submitted 31 May, 2023; v1 submitted 1 August, 2022; originally announced August 2022.

    Journal ref: Proceedings of the 40th International Conference on Machine Learning, PMLR 202. 2023

  11. arXiv:2206.12314  [pdf, other

    stat.ML cs.LG

    Learning sparse features can lead to overfitting in neural networks

    Authors: Leonardo Petrini, Francesco Cagnetta, Eric Vanden-Eijnden, Matthieu Wyart

    Abstract: It is widely believed that the success of deep networks lies in their ability to learn a meaningful representation of the features of the data. Yet, understanding when and how this feature learning improves performance remains a challenge: for example, it is beneficial for modern architectures trained to classify images, whereas it is detrimental for fully-connected networks trained for the same t… ▽ More

    Submitted 12 October, 2022; v1 submitted 24 June, 2022; originally announced June 2022.

  12. arXiv:2106.08619  [pdf, other

    stat.ML cond-mat.dis-nn cs.LG

    Locality defeats the curse of dimensionality in convolutional teacher-student scenarios

    Authors: Alessandro Favero, Francesco Cagnetta, Matthieu Wyart

    Abstract: Convolutional neural networks perform a local and translationally-invariant treatment of the data: quantifying which of these two aspects is central to their success remains a challenge. We study this problem within a teacher-student framework for kernel regression, using `convolutional' kernels inspired by the neural tangent kernel of simple convolutional architectures of given filter size. Using… ▽ More

    Submitted 12 November, 2021; v1 submitted 16 June, 2021; originally announced June 2021.

    Comments: 32 pages, 7 figures

  13. arXiv:2104.06764  [pdf, other

    cond-mat.stat-mech cond-mat.soft

    Universal properties of active membranes

    Authors: Francesco Cagnetta, Viktor Skultety, Martin R. Evans, Davide Marenduzzo

    Abstract: We put forward a general field theory for membranes with embedded activators and analyse their critical properties using renormalization group techniques. Depending on the membrane-activator coupling, we find a crossover between acoustic and diffusive scaling regimes, with mean-field dynamical critical exponents z = 1 and 2 respectively. We argue that the acoustic scaling, which is exact in all sp… ▽ More

    Submitted 16 November, 2021; v1 submitted 14 April, 2021; originally announced April 2021.

    Comments: 5 pages, 3 figures

  14. arXiv:2104.06762  [pdf, other

    cond-mat.stat-mech cond-mat.soft

    A renormalization group study of the dynamics of active membranes: universality classes and scaling laws

    Authors: Francesco Cagnetta, Viktor Skultety, Martin R. Evans, Davide Marenduzzo

    Abstract: Motivated by experimental observations of patterning at the leading edge of motile eukaryotic cells, we introduce a general model for the dynamics of nearly-flat fluid membranes driven from within by an ensemble of activators. We include, in particular, a kinematic coupling between activator density and membrane slope which generically arises whenever the membrane has a non-vanishing normal speed.… ▽ More

    Submitted 16 November, 2021; v1 submitted 14 April, 2021; originally announced April 2021.

  15. arXiv:2103.11369  [pdf, other

    cond-mat.stat-mech

    Work Fluctuations in the Active Ornstein- Uhlenbeck Particle model

    Authors: Massimiliano Semeraro, Antonio Suma, Isabella Petrelli, Francesco Cagnetta, Giuseppe Gonnella

    Abstract: We study the large deviations of the power injected by the active force for an Active Ornstein-Uhlenbeck Particle (AOUP), free or in a confining potential. For the free-particle case, we compute the rate function analytically in d-dimensions from a saddle-point expansion, and numerically in two dimensions by it a) direct sampling of the active work in numerical solutions of the AOUP equations and… ▽ More

    Submitted 30 September, 2021; v1 submitted 21 March, 2021; originally announced March 2021.

    Comments: 33 pages, 12 figures

  16. arXiv:2001.08500  [pdf, other

    cond-mat.stat-mech cond-mat.soft

    Work fluctuations of self-propelled particles in the phase separated state

    Authors: P. Chiarantoni, F. Cagnetta, F. Corberi, G. Gonnella, A. Suma

    Abstract: We study the large deviations of the distribution P(W_τ) of the work associated with the propulsion of individual active brownian particles in a time interval τ, in the region of the phase diagram where macroscopic phase separation takes place. P(W_τ) is characterised by two peaks, associated to particles in the gaseous and in the clusterised phases, and two separate non-convex branches. According… ▽ More

    Submitted 18 May, 2020; v1 submitted 23 January, 2020; originally announced January 2020.

    Comments: 7 pages, 5 figures

  17. arXiv:1912.07281  [pdf, ps, other

    cond-mat.stat-mech cond-mat.soft

    Kinetic roughening in active interfaces

    Authors: Francesco Cagnetta, Martin R. Evans, Davide Marenduzzo

    Abstract: The essential features of many interfaces driven out of equilibrium are described by the same equation---the Kardar-Parisi-Zhang (KPZ) equation. How do living interfaces, such as the cell membrane, fit into this picture? In an endeavour to answer such a question, we proposed in [F. Cagnetta, M. R. Evans, D. Marenduzzo, PRL 120, 258001 (2018)] an idealised model for the membrane of a moving cell. H… ▽ More

    Submitted 16 December, 2019; originally announced December 2019.

    Comments: 5 pages, 4 figures, FisMat 2019

  18. arXiv:1908.06671  [pdf, other

    q-bio.SC cond-mat.soft physics.bio-ph

    A nonequilibrium strategy for fast target search on the genome

    Authors: F. Cagnetta, D. Michieletto, D. Marenduzzo

    Abstract: Vital biological processes such as genome repair require fast and efficient binding of selected proteins to specific target sites on DNA. Here we propose an active target search mechanism based on "chromophoresis", the dynamics of DNA-binding proteins up or down gradients in the density of epigenetic marks, or colours (biochemical tags on the genome). We focus on a set of proteins that deposit mar… ▽ More

    Submitted 23 April, 2020; v1 submitted 19 August, 2019; originally announced August 2019.

    Comments: 5 pages, 5 figures

    Journal ref: Phys. Rev. Lett. 124, 198101 (2020)

  19. Efficiency of one-dimensional active transport conditioned on motility

    Authors: Francesco Cagnetta, Emil Mallmin

    Abstract: By conditioning a stochastic process on the value of an observable, one obtains a new stochastic process with different properties. We apply this idea in the context of active matter, and condition interacting self-propelled particles on their individual motility. Using the effective process formalism from dynamical large deviations theory, we derive the interactions that actuate the imposed mobil… ▽ More

    Submitted 12 February, 2020; v1 submitted 2 July, 2019; originally announced July 2019.

    Comments: 11 pages, 9 figures

    Journal ref: Phys. Rev. E 101, 022130 (2020)

  20. arXiv:1904.08985  [pdf, other

    cond-mat.stat-mech

    Inviscid limit of the active interface equations

    Authors: Francesco Cagnetta, Martin R. Evans

    Abstract: We present a detailed solution of the active interface equations in the inviscid limit. The active interface equations were previously introduced as a toy model of membrane-protein systems: they describe a stochastic interface where growth is stimulated by inclusions which themselves move on the interface. In the inviscid limit, the equations reduce to a pair of coupled conservation laws. After di… ▽ More

    Submitted 28 August, 2019; v1 submitted 18 April, 2019; originally announced April 2019.

    Comments: 22 pages, 11 figures

    Journal ref: J. Stat. Mech. (2019) 113206

  21. arXiv:1811.12903  [pdf, ps, other

    cond-mat.stat-mech

    Statistical mechanics of a single active slider on a fluctuating interface

    Authors: Francesco Cagnetta, Martin R. Evans, Davide Marenduzzo

    Abstract: We study the statistical mechanics of a single active slider on a fluctuating interface, by means of numerical simulations and theoretical arguments. The slider, which moves by definition towards the interface minima, is active as it also stimulates growth of the interface. Even though such a particle has no counterpart in thermodynamic systems, active sliders may provide a simple model for ATP-de… ▽ More

    Submitted 17 April, 2019; v1 submitted 30 November, 2018; originally announced November 2018.

    Comments: 13 pages, 19 figures

    Journal ref: Phys. Rev. E 99, 042124 (2019)

  22. arXiv:1712.05764  [pdf, ps, other

    cond-mat.soft cond-mat.stat-mech physics.bio-ph

    Active interface growth and pattern formation in membrane-protein systems

    Authors: F. Cagnetta, M. R. Evans, D. Marenduzzo

    Abstract: Inspired by recent experimental observation of patterning at the membrane of a living cell, we propose a generic model for the dynamics of a fluctuating interface driven by particle-like inclusions which stimulate its growth. We find that the coupling between interfacial and inclusions dynam- ics yields microphase separation and the self-organisation of travelling waves. These patterns are strikin… ▽ More

    Submitted 11 May, 2018; v1 submitted 15 December, 2017; originally announced December 2017.

    Comments: 5 pages, 5 figures

    Journal ref: Phys. Rev. Lett. 120, 258001 (2018)

  23. Large fluctuations and dynamic phase transition in a system of self-propelled particles

    Authors: Francesco Cagnetta, Federico Corberi, Giuseppe Gonnella, Antonio Suma

    Abstract: We study the statistics, in stationary conditions, of the work $W_τ$ done by the active force in different systems of self-propelled particles in a time $τ$. We show the existence of a critical value $W_τ^†$ such that fluctuations with $W_τ>W_τ^†$ correspond to configurations where interaction between particles plays a minor role whereas those with $W_τ< W_τ^†$ represent states with single particl… ▽ More

    Submitted 15 October, 2017; originally announced October 2017.

    Comments: 6 pages, 4 figures

    Journal ref: Phys. Rev. Lett. 119, 158002 (2017)

  24. arXiv:1504.05357  [pdf, other

    nlin.CD cond-mat.stat-mech

    Strong anomalous diffusion of the phase of a chaotic pendulum

    Authors: Francesco Cagnetta, Giuseppe Gonnella, Alessandro Mossa, Stefano Ruffo

    Abstract: In this letter we consider the phase diffusion of a harmonically driven undamped pendulum and show that it is anomalous in the strong sense. The role played by the fractal properties of the phase space is highlighted, providing an illustration of the link between deterministic chaos and anomalous transport. Finally, we build a stochastic model which reproduces most properties of the original Hamil… ▽ More

    Submitted 21 April, 2015; originally announced April 2015.

    Comments: 6 pages, 6 figures

    MSC Class: 70K55; 82C70; 34C28

    Journal ref: Europhys. Lett., 111 (2015) 10002