Archivio istituzionale della Ricerca - Scuola Normale Superiore
Not a member yet
27040 research outputs found
Sort by
The Blown Head Gasket Effect: Rise and Struggles of The Drivers Cooperative in New York City
No full textThis paper presents an in-depth case study of The Drivers Cooperative (TDC), a driver-owned ride-hailing platform based in New York City, to examine the challenges faced by multistakeholder platform cooperatives. Drawing on 16 semi-structured interviews, six months of participant observation, and extensive desk research, the study traces how TDC’s growth efforts ultimately led to a fracture between its cooperative membership and its platform infrastructure. This dynamic – conceptualised as the Blown Head Gasket Effect – illustrates how diverging stakeholders can fracture during the development of democratic platform alternatives. Factors such as limited access to capital, political divergence among stakeholders, and temporal pressures are identified as key drivers of this effect. The paper concludes by discussing TDC’s legislative achievements and offering practical recommendations for cooperative practitioners and policy interventions. By situating TDC within broader debates on platform cooperativism and the digital solidarity economy, the study underscores the importance of learning from defeats to enhance cooperative business resilience
Scultura monumentale ad Akrai in età arcaica. Un riesame
No full textThis paper investigates two life-size limestone statues from Akrai, Syracuse’s second-generation apoikia, that are generally dated by scholarship between the end of the 7th and the early 6th century BC. However, despite their historical-artistic relevance, the two sculptures have never been systematically investigated, also due to their poor conditions. Therefore, this contribution aims to bridge this gap, by providing both sculptures with a new chronological and art historical framework, in light of Akrai’s archaeological context
Explaining Machine Learning and Memorization with Statistical Mechanics
Full text linkArtificial neural networks (NNs) and machine learning (ML) algorithms are poorly understood from a theoretical perspective, which makes it difficult to fully realize their potential and overcome their weaknesses. For instance, ML algorithms train NN weights by moving them along a low-dimensional subspace of their allowed values, but this implicitly low-dimensional learning structure is not properly exploited to improve training because its nature is not well understood. Moreover, trained NNs are easily confused by pervasive adversarial attacks whose theoretical underpinnings are still unclear. This thesis aims to improve our theoretical understanding of NNs and ML, with a particular focus on adversarial attacks and implicitly low-dimensional learning. For this purpose, we use mathematical tools from statistical mechanics to study different types of NNs and ways in which they can fit the data. In particular, we study two classes of models that fit the data with various degrees of learning and memorization: dense associative memory (DAM) and restricted Boltzmann machines (RBM). In the process, we investigate connections between different versions of these models that are useful to make analytical investigations more efficient.
First, we study a type of DAM called dense Hopfield network (dense HN) in the teacher-student setting where it is trained using data generated by another dense HN. On the Nishimori line, we show that the phase where dense HNs in the teacher-student setting are able to learn data coincides with the spin-glass phase of dense HNs with random memorized patterns. Outside the Nishimori line, we investigate the noise tolerance and adversarial robustness of dense HNs. In particular, we derive an exact formula for the adversarial robustness of the student at zero temperature, and we clarify why the adversarial robustness of dense HNs changes as a function of the learning regime.
Second, we study RBMs in the teacher-student setting. When the teacher's weights are uncorrelated, we validate the conjecture that the performance of the student in learning them is independent of the number of hidden units. Moreover, we show that a student that is larger than necessary to learn the teacher's weights adopts a low-dimensional learning strategy in which only a subset of its hidden units end up correlated with those of the teacher, which we argue can be used as a toy model for studying the lottery ticket hypothesis. When the teacher's weights are correlated together rather than purely random, we show that the student crosses multiple regimes of data representation where it learns them in increasingly detailed ways as the number of samples in its training dataset increases.
Finally, we study a type of RBM that belongs to the class of DAMs and is capable of both supervised and unsupervised classification. As before, our methods are based on statistical mechanics calculations in the teacher-student setting. We propose a novel regularization scheme inspired by these calculations, which we find to make training on real data significantly more stable. Moreover, we show that the weights learned by relatively small DAMs trained on both real and synthetic data are saddle points of larger DAMs, and we implement an algorithm that uses this hierarchy to significantly accelerate training on real data
Vertical excitation energies of embedded systems: The vertical excitation model (VEM) within polarizable QM/MM
No full textL'imperfetto narrativo in greco: registro, genere letterario e diacronia
Full text linkNelle opere narrative greche antiche, così come nelle opere narrative composte in molte altre lingue, le forme di passato perfettivo aoristico (gli indicativi aoristi) sono usate per far progredire la narrazione. Esse denotano sequenze di eventi in primo piano ordinati dal punto di vista temporale. Le forme di passato imperfettivo (gli imperfetti) sono usate, invece, per designare eventi in secondo piano, come descrizioni o commenti dell'autore. Tuttavia, in greco antico, le forme di passato imperfettivo possono essere impiegate in contesti perfettivi per far progredire la narrazione della sequenza principale di eventi. Tali forme di imperfetto sono chiamate "imperfetti narrativi". Lo scopo di questa tesi è fornire una descrizione dello sviluppo diacronico dell'imperfetto narrativo nelle opere letterarie greche. La tesi mira, inoltre, a costruire un corpus esaustivo di diversi testi narrativi greci, con lo scopo di osservare come gli imperfetti narrativi siano utilizzati nei testi letterari greci antichi e bizantini. I testi scelti sono: il primo libro delle Storie di Tucidide; il primo libro delle Storie di Polibio; il primo libro delle Antichità romane di Dionigi di Alicarnasso; il Vangelo di Luca; la Storia lausiaca di Palladio; l'Historia monachorum in Aegypto; il primo libro della Storia di Procopio di Cesarea; i primi due libri delle Storie di Agazia; i primi due libri dell'Alessiade di Anna Comnena; i versi da 1 a 700 delle redazioni H e P della Cronaca della Morea. I testi scelti sono diversi per epoca e registro: questa scelta è stata operata in modo da osservare se l'epoca di composizione e il registro delle opere interagiscono con l'uso dell'imperfetto narrativo. Nelle opere tardoantiche e bizantine analizzate, l'uso dell'imperfetto narrativo sembra interagire con il registro dei testi. Inoltre, in tutti i periodi presi in considerazione in questa tesi, esiste una forte connessione tra il parametro azionale della duratività e l'uso di forme di imperfetto narrativo.In Ancient Greek narrative texts, as well as in other languages, aoristic past forms (aorist indicatives) are used to "push forward" the narration. They are used to denote temporally sequential foregrounded events. On the contrary, imperfective past forms (imperfects) are used to denote backgrounded events, such as descriptions or comments made by the author. However, in Ancient Greek, imperfective past forms can be used in perfective contexts to "push forward" the narration of the main sequence of events. Such an imperfect form is called "narrative imperfect". The aim of this thesis is to provide a description of the diachronic development of the narrative imperfect in Greek literary texts and to build a comprehensive corpus of several narrative literary texts in order to investigate how narrative imperfects are used in both Ancient and Byzantine Greek narrative texts. The texts which have been chosen are: the first book of Thucydides' Histories, the first book of Polybius' Histories, the first book of Dionysius' Roman Antiquities, the Gospel of Luke, Palladius' Historia Lausiaca, the Historia monachorum in Aegypto, the first book of Procopius' Wars, the first two books of Agathias' Histories, the first two books of Anna Comnene's Alexiad, and the first 700 verses of the Chronicle of Morea (version H and version P). The texts which have been chosen are different by age and register, in order to investigate whether the age of composition and the register of the text interact with the use of the narrative imperfect. In Late-Antique and Byzantine Greek, the use of narrative imperfect forms seems to interact with the register of the texts. In all the periods that are under investigation in this thesis, there is also a strong connection between the actional parameter of durativity and the use of narrative imperfect forms, with durative verbs using the narrative imperfect much more frequently than punctual verbs
The Stefan Problem with Mushy Region as a Scaling Limit of Stochastic PDE with Turbulent Transport
No full textFalling behind unequally : labour market outcomes of Italian couples after childbirth*
No full textThis study explores how childbirth differently shapes the career trajectories of men and women within the same couples, with a particular focus on gender disparities in experiencing downward labour transitions following the birth of their first child. Using a unique survey-administrative linked dataset, we track couples’labour market trajectories to analyse transitions from employment to unemployment, full-time to part-time employment, and higher-paid to lower-paid jobs. Additionally, the dataset allows to link partners, enabling the study of factors influencing differences in the probabilities of downward labour market transitions between partners in the same household. Our findings reveal substantial and persistent penalties for women, lasting up to three years after childbirth, which are mainly related to part-time job arrangements. When examining differences in probabilities within couples, households in which women have tertiary education with respect to their partners and are the primary earners exhibit smaller gender disparities in the likelihood of downward labour transitions with respect to other households
