@unpublished{20572,
  abstract     = {We present an elementary non-recursive formula for the multivariate moments
of the Dirichlet distribution on the standard simplex, in terms of the pattern
inventory of the moments' exponents. We obtain analog formulas for the
multivariate moments of the Dirichlet-Ferguson and Gamma measures. We further
introduce a polychromatic analogue of Ewens sampling formula on colored integer
partitions, discuss its relation with suitable extensions of Hoppe's urn model
and of the Chinese restaurant process, and prove that it satisfies an adapted
notion of consistency in the sense of Kingman.},
  author       = {Dello Schiavo, Lorenzo and Quattrocchi, Filippo},
  booktitle    = {arXiv},
  keywords     = {Dirichlet distribution, Ewens sampling formula, Hoppe urn model, colored partitions},
  title        = {{Multivariate Dirichlet moments and a polychromatic Ewens sampling formula}},
  doi          = {10.48550/arXiv.2309.11292},
  year         = {2023},
}

@inproceedings{14863,
  author       = {Polesello, Andrea and Muller, Caroline J and Pasquero, Claudia and Meroni, Agostino N.},
  booktitle    = {EGU General Assembly 2023},
  location     = {Vienna, Austria & Virtual},
  publisher    = {European Geosciences Union},
  title        = {{Intensification mechanisms of tropical cyclones}},
  doi          = {10.5194/egusphere-egu23-6157},
  year         = {2023},
}

@phdthesis{12885,
  abstract     = {High-performance semiconductors rely upon precise control of heat and charge transport. This can be achieved by precisely engineering defects in polycrystalline solids. There are multiple approaches to preparing such polycrystalline semiconductors, and the transformation of solution-processed colloidal nanoparticles is appealing because colloidal nanoparticles combine low cost with structural and compositional tunability along with rich surface chemistry. However, the multiple processes from nanoparticle synthesis to the final bulk nanocomposites are very complex. They involve nanoparticle purification, post-synthetic modifications, and finally consolidation (thermal treatments and densification). All these properties dictate the final material’s composition and microstructure, ultimately affecting its functional properties. This thesis explores the synthesis, surface chemistry and consolidation of colloidal semiconductor nanoparticles into dense solids. In particular, the transformations that take place during these processes, and their effect on the material’s transport properties are evaluated. },
  author       = {Calcabrini, Mariano},
  isbn         = {978-3-99078-028-2},
  issn         = {2663-337X},
  pages        = {82},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Nanoparticle-based semiconductor solids: From synthesis to consolidation}},
  doi          = {10.15479/at:ista:12885},
  year         = {2023},
}

@inproceedings{14260,
  abstract     = {This paper presents Lincheck, a new practical and user-friendly framework for testing concurrent algorithms on the Java Virtual Machine (JVM). Lincheck provides a simple and declarative way to write concurrent tests: instead of describing how to perform the test, users specify what to test by declaring all the operations to examine; the framework automatically handles the rest. As a result, tests written with Lincheck are concise and easy to understand. The framework automatically generates a set of concurrent scenarios, examines them using stress-testing or bounded model checking, and verifies that the results of each invocation are correct. Notably, if an error is detected via model checking, Lincheck provides an easy-to-follow trace to reproduce it, significantly simplifying the bug investigation.

To the best of our knowledge, Lincheck is the first production-ready tool on the JVM that offers such a simple way of writing concurrent tests, without requiring special skills or expertise. We successfully integrated Lincheck in the development process of several large projects, such as Kotlin Coroutines, and identified new bugs in popular concurrency libraries, such as a race in Java’s standard ConcurrentLinkedDeque and a liveliness bug in Java’s AbstractQueuedSynchronizer framework, which is used in most of the synchronization primitives. We believe that Lincheck can significantly improve the quality and productivity of concurrent algorithms research and development and become the state-of-the-art tool for checking their correctness.},
  author       = {Koval, Nikita and Fedorov, Alexander and Sokolova, Maria and Tsitelov, Dmitry and Alistarh, Dan-Adrian},
  booktitle    = {35th International Conference on Computer Aided Verification},
  isbn         = {9783031377051},
  issn         = {1611-3349},
  location     = {Paris, France},
  pages        = {156--169},
  publisher    = {Springer Nature},
  title        = {{Lincheck: A practical framework for testing concurrent data structures on JVM}},
  doi          = {10.1007/978-3-031-37706-8_8},
  volume       = {13964},
  year         = {2023},
}

@phdthesis{14547,
  abstract     = {Superconductor-semiconductor heterostructures currently capture a significant amount of research interest and they serve as the physical platform in many proposals towards topological quantum computation.
Despite being under extensive investigations, historically using transport techniques, the basic properties of the interface between the superconductor and the semiconductor remain to be understood.

In this thesis, two separate studies on the Al-InAs heterostructures are reported with the first focusing on the physics of the material motivated by the emergence of a new phase, the Bogoliubov-Fermi surface. 
The second focuses on a technological application, a gate-tunable Josephson parametric amplifier.

In the first study, we investigate the hypothesized unconventional nature of the induced superconductivity at the interface between the Al thin film and the InAs quantum well.
We embed a two-dimensional Al-InAs hybrid system in a resonant microwave circuit allowing measurements of change in inductance.
The behaviour of the resonance in a range of temperature and in-plane magnetic field has been studied and compared with the theory of conventional s-wave superconductor and a two-component theory that includes both contribution of the $s$-wave pairing in Al and the intraband $p \pm ip$ pairing in InAs.
Measuring the temperature dependence of resonant frequency, no discrepancy is found between data and the conventional theory.
We observe the breakdown of superconductivity due to an applied magnetic field which contradicts the conventional theory.
In contrast, the data can be captured quantitatively by fitting to a two-component model.
We find the evidence of the intraband $p \pm ip$ pairing in the InAs and the emergence of the Bogoliubov-Fermi surfaces due to magnetic field with the characteristic value $B^* = 0.33~\mathrm{T}$.
From the fits, the sheet resistance of Al, the carrier density and mobility in InAs are determined.
By systematically studying the anisotropy of the circuit response, we find weak anisotropy for $B < B^*$ and increasingly strong anisotropy for $B > B^*$ resulting in a pronounced two-lobe structure in polar plot of frequency versus field angle.
Strong resemblance between the field dependence of dissipation and superfluid density hints at a hidden signature of the Bogoliubov-Fermi surface that is burried in the dissipation data.

In the second study, we realize a parametric amplifier with a Josephson field effect transistor as the active element.
The device's modest construction consists of a gated SNS weak link embedded at the center of a coplanar waveguide resonator.
By applying a gate voltage, the resonant frequency is field-effect tunable over a range of 2 GHz.
Modelling the JoFET minimally as a parallel RL circuit, the dissipation introduced by the JoFET can be quantitatively related to the gate voltage.
We observed gate-tunable Kerr nonlinearity qualitatively in line with expectation.
The JoFET amplifier has 20 dB of gain, 4 MHz of instantaneous bandwidth, and a 1dB compression point of -125.5 dBm when operated at a fixed resonant frequency.
In general, the signal-to-noise ratio is improved by 5-7 dB when the JoFET amplifier is activated compared.
The noise of the measurement chain and insertion loss of relevant circuit elements are calibrated to determine the expected and the real noise performance of the JoFET amplifier.
As a quantification of the noise performance, the measured total input-referred noise of the JoFET amplifier is in good agreement with the estimated expectation which takes device loss into account.
We found that the noise performance of the device reported in this document approaches one photon of total input-referred added noise which is the quantum limit imposed in nondegenerate parametric amplifier.},
  author       = {Phan, Duc T},
  issn         = {2663-337X},
  keywords     = {superconductor-semiconductor, superconductivity, Al, InAs, p-wave, superconductivity, JPA, microwave},
  pages        = {80},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Resonant microwave spectroscopy of Al-InAs}},
  doi          = {10.15479/14547},
  year         = {2023},
}

@article{12675,
  abstract     = {Aromatic side chains are important reporters of the plasticity of proteins, and often form important contacts in protein--protein interactions. By studying a pair of structurally homologous cross-β amyloid fibrils, HET-s and HELLF, with a specific isotope-labeling approach and magic-angle-spinning (MAS) NMR, we have characterized the dynamic behavior of Phe and Tyr aromatic rings to show that the hydrophobic amyloid core is rigid, without any sign of "breathing motions" over hundreds of milliseconds at least. Aromatic residues exposed at the fibril surface have a rigid ring axis but undergo ring flips, on a variety of time scales from ns to µs. Our approach provides direct insight into hydrophobic-core motions, enabling a better evaluation of the conformational heterogeneity generated from a NMR structural ensemble of such amyloid cross-β architecture.},
  author       = {Becker, Lea Marie and Berbon, Mélanie and Vallet, Alicia and Grelard, Axelle and Morvan, Estelle and Bardiaux, Benjamin and Lichtenecker, Roman and Ernst, Matthias and Loquet, Antoine and Schanda, Paul},
  issn         = {1521-3773},
  journal      = {Angewandte Chemie International Edition},
  keywords     = {General Chemistry, Catalysis},
  number       = {19},
  publisher    = {Wiley},
  title        = {{The rigid core and flexible surface of amyloid fibrils probed by Magic‐Angle Spinning NMR of aromatic residues}},
  doi          = {10.1002/anie.202219314},
  volume       = {62},
  year         = {2023},
}

@unpublished{15039,
  abstract     = {A crucial property for achieving secure, trustworthy and interpretable deep learning systems is their robustness: small changes to a system's inputs should not result in large changes to its outputs. Mathematically, this means one strives for networks with a small Lipschitz constant. Several recent works have focused on how to construct such Lipschitz networks, typically by imposing constraints on the weight matrices. In this work, we study an orthogonal aspect, namely the role of the activation function. We show that commonly used activation functions, such as MaxMin, as well as all piece-wise linear ones with two segments unnecessarily restrict the class of representable functions, even in the simplest one-dimensional setting. We furthermore introduce the new N-activation function that is provably more expressive than currently popular activation functions. We provide code at this https URL.},
  author       = {Prach, Bernd and Lampert, Christoph},
  booktitle    = {arXiv},
  title        = {{1-Lipschitz neural networks are more expressive with N-activations}},
  doi          = {10.48550/ARXIV.2311.06103},
  year         = {2023},
}

@article{12762,
  abstract     = {Neurons in the brain are wired into adaptive networks that exhibit collective dynamics as diverse as scale-specific oscillations and scale-free neuronal avalanches. Although existing models account for oscillations and avalanches separately, they typically do not explain both phenomena, are too complex to analyze analytically or intractable to infer from data rigorously. Here we propose a feedback-driven Ising-like class of neural networks that captures avalanches and oscillations simultaneously and quantitatively. In the simplest yet fully microscopic model version, we can analytically compute the phase diagram and make direct contact with human brain resting-state activity recordings via tractable inference of the model’s two essential parameters. The inferred model quantitatively captures the dynamics over a broad range of scales, from single sensor oscillations to collective behaviors of extreme events and neuronal avalanches. Importantly, the inferred parameters indicate that the co-existence of scale-specific (oscillations) and scale-free (avalanches) dynamics occurs close to a non-equilibrium critical point at the onset of self-sustained oscillations.},
  author       = {Lombardi, Fabrizio and Pepic, Selver and Shriki, Oren and Tkačik, Gašper and De Martino, Daniele},
  issn         = {2662-8457},
  journal      = {Nature Computational Science},
  pages        = {254--263},
  publisher    = {Springer Nature},
  title        = {{Statistical modeling of adaptive neural networks explains co-existence of avalanches and oscillations in resting human brain}},
  doi          = {10.1038/s43588-023-00410-9},
  volume       = {3},
  year         = {2023},
}

@inproceedings{13120,
  abstract     = {We formalized general (i.e., type-0) grammars using the Lean 3 proof assistant. We defined basic notions of rewrite rules and of words derived by a grammar, and used grammars to show closure of the class of type-0 languages under four operations: union, reversal, concatenation, and the Kleene star. The literature mostly focuses on Turing machine arguments, which are possibly more difficult to formalize. For the Kleene star, we could not follow the literature and came up with our own grammar-based construction.},
  author       = {Dvorak, Martin and Blanchette, Jasmin},
  booktitle    = {14th International Conference on Interactive Theorem Proving},
  isbn         = {9783959772846},
  issn         = {1868-8969},
  location     = {Bialystok, Poland},
  publisher    = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik},
  title        = {{Closure properties of general grammars - formally verified}},
  doi          = {10.4230/LIPIcs.ITP.2023.15},
  volume       = {268},
  year         = {2023},
}

@article{13049,
  abstract     = {We propose a computational design approach for covering a surface with individually addressable RGB LEDs, effectively forming a low-resolution surface screen. To achieve a low-cost and scalable approach, we propose creating designs from flat PCB panels bent in-place along the surface of a 3D printed core. Working with standard rigid PCBs enables the use of
established PCB manufacturing services, allowing the fabrication of designs with several hundred LEDs. 
Our approach optimizes the PCB geometry for folding, and then jointly optimizes the LED packing, circuit and routing, solving a challenging layout problem under strict manufacturing requirements. Unlike paper, PCBs cannot bend beyond a certain point without breaking. Therefore, we introduce parametric cut patterns acting as hinges, designed to allow bending while remaining compact. To tackle the joint optimization of placement, circuit and routing, we propose a specialized algorithm that splits the global problem into one sub-problem per triangle, which is then individually solved.
Our technique generates PCB blueprints in a completely automated way. After being fabricated by a PCB manufacturing service, the boards are bent and glued by the user onto the 3D printed support. We demonstrate our technique on a range of physical models and virtual examples, creating intricate surface light patterns from hundreds of LEDs.},
  author       = {Freire, Marco and Bhargava, Manas and Schreck, Camille and Hugron, Pierre-Alexandre and Bickel, Bernd and Lefebvre, Sylvain},
  issn         = {1557-7368},
  journal      = {Transactions on Graphics},
  keywords     = {PCB design and layout, Mesh geometry models},
  location     = {Los Angeles, CA, United States},
  number       = {4},
  publisher    = {Association for Computing Machinery},
  title        = {{PCBend: Light up your 3D shapes with foldable circuit boards}},
  doi          = {10.1145/3592411},
  volume       = {42},
  year         = {2023},
}

@article{14343,
  abstract     = {The total energy of an eigenstate in a composite quantum system tends to be distributed equally among its constituents. We identify the quantum fluctuation around this equipartition principle in the simplest disordered quantum system consisting of linear combinations of Wigner matrices. As our main ingredient, we prove the Eigenstate Thermalisation Hypothesis and Gaussian fluctuation for general quadratic forms of the bulk eigenvectors of Wigner matrices with an arbitrary deformation.},
  author       = {Cipolloni, Giorgio and Erdös, László and Henheik, Sven Joscha and Kolupaiev, Oleksii},
  issn         = {2050-5094},
  journal      = {Forum of Mathematics, Sigma},
  publisher    = {Cambridge University Press},
  title        = {{Gaussian fluctuations in the equipartition principle for Wigner matrices}},
  doi          = {10.1017/fms.2023.70},
  volume       = {11},
  year         = {2023},
}

@article{14421,
  abstract     = {Only recently has it been possible to construct a self-adjoint Hamiltonian that involves the creation of Dirac particles at a point source in 3d space. Its definition makes use of an interior-boundary condition. Here, we develop for this Hamiltonian a corresponding theory of the Bohmian configuration. That is, we (non-rigorously) construct a Markov jump process $(Q_t)_{t\in\mathbb{R}}$ in the configuration space of a variable number of particles that is $|\psi_t|^2$-distributed at every time t and follows Bohmian trajectories between the jumps. The jumps correspond to particle creation or annihilation events and occur either to or from a configuration with a particle located at the source. The process is the natural analog of Bell's jump process, and a central piece in its construction is the determination of the rate of particle creation. The construction requires an analysis of the asymptotic behavior of the Bohmian trajectories near the source. We find that the particle reaches the source with radial speed 0, but orbits around the source infinitely many times in finite time before absorption (or after emission).},
  author       = {Henheik, Sven Joscha and Tumulka, Roderich},
  issn         = {1751-8121},
  journal      = {Journal of Physics A: Mathematical and Theoretical},
  number       = {44},
  publisher    = {IOP Publishing},
  title        = {{Creation rate of Dirac particles at a point source}},
  doi          = {10.1088/1751-8121/acfe62},
  volume       = {56},
  year         = {2023},
}

@article{14425,
  abstract     = {Water adsorption and dissociation processes on pristine low-index TiO2 interfaces are important but poorly understood outside the well-studied anatase (101) and rutile (110). To understand these, we construct three sets of machine learning potentials that are simultaneously applicable to various TiO2 surfaces, based on three density-functional-theory approximations. Here we show the water dissociation free energies on seven pristine TiO2 surfaces, and predict that anatase (100), anatase (110), rutile (001), and rutile (011) favor water dissociation, anatase (101) and rutile (100) have mostly molecular adsorption, while the simulations of rutile (110) sensitively depend on the slab thickness and molecular adsorption is preferred with thick slabs. Moreover, using an automated algorithm, we reveal that these surfaces follow different types of atomistic mechanisms for proton transfer and water dissociation: one-step, two-step, or both. These mechanisms can be rationalized based on the arrangements of water molecules on the different surfaces. Our finding thus demonstrates that the different pristine TiO2 surfaces react with water in distinct ways, and cannot be represented using just the low-energy anatase (101) and rutile (110) surfaces.},
  author       = {Zeng, Zezhu and Wodaczek, Felix and Liu, Keyang and Stein, Frederick and Hutter, Jürg and Chen, Ji and Cheng, Bingqing},
  issn         = {2041-1723},
  journal      = {Nature Communications},
  publisher    = {Springer Nature},
  title        = {{Mechanistic insight on water dissociation on pristine low-index TiO2 surfaces from machine learning molecular dynamics simulations}},
  doi          = {10.1038/s41467-023-41865-8},
  volume       = {14},
  year         = {2023},
}

@phdthesis{13074,
  abstract     = {Deep learning has become an integral part of a large number of important applications, and many of the recent breakthroughs have been enabled by the ability to train very large models, capable to capture complex patterns and relationships from the data. At the same time, the massive sizes of modern deep learning models have made their deployment to smaller devices more challenging; this is particularly important, as in many applications the users rely on accurate deep learning predictions, but they only have access to devices with limited memory and compute power. One solution to this problem is to prune neural networks, by setting as many of their parameters as possible to zero, to obtain accurate sparse models with lower memory footprint. Despite the great research progress in obtaining sparse models that preserve accuracy, while satisfying memory and computational constraints, there are still many challenges associated with efficiently training sparse models, as well as understanding their generalization properties.

The focus of this thesis is to investigate how the training process of sparse models can be made more efficient, and to understand the differences between sparse and dense models in terms of how well they can generalize to changes in the data distribution. We first study a method for co-training sparse and dense models, at a lower cost compared to regular training. With our method we can obtain very accurate sparse networks, and dense models that can recover the baseline accuracy. Furthermore, we are able to more easily analyze the differences, at prediction level, between the sparse-dense model pairs. Next, we investigate the generalization properties of sparse neural networks in more detail, by studying how well different sparse models trained on a larger task can adapt to smaller, more specialized tasks, in a transfer learning scenario. Our analysis across multiple pruning methods and sparsity levels reveals that sparse models provide features that can transfer similarly to or better than the dense baseline. However, the choice of the pruning method plays an important role, and can influence the results when the features are fixed (linear finetuning), or when they are allowed to adapt to the new task (full finetuning). Using sparse models with fixed masks for finetuning on new tasks has an important practical advantage, as it enables training neural networks on smaller devices. However, one drawback of current pruning methods is that the entire training cycle has to be repeated to obtain the initial sparse model, for every sparsity target; in consequence, the entire training process is costly and also multiple models need to be stored. In the last part of the thesis we propose a method that can train accurate dense models that are compressible in a single step, to multiple sparsity levels, without additional finetuning. Our method results in sparse models that can be competitive with existing pruning methods, and which can also successfully generalize to new tasks.},
  author       = {Peste, Elena-Alexandra},
  issn         = {2663-337X},
  pages        = {147},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Efficiency and generalization of sparse neural networks}},
  doi          = {10.15479/at:ista:13074},
  year         = {2023},
}

@inproceedings{13221,
  abstract     = {The safety-liveness dichotomy is a fundamental concept in formal languages which plays a key role in verification. Recently, this dichotomy has been lifted to quantitative properties, which are arbitrary functions from infinite words to partially-ordered domains. We look into harnessing the dichotomy for the specific classes of quantitative properties expressed by quantitative automata. These automata contain finitely many states and rational-valued transition weights, and their common value functions Inf, Sup, LimInf, LimSup, LimInfAvg, LimSupAvg, and DSum map infinite words into the totallyordered domain of real numbers. In this automata-theoretic setting, we establish a connection between quantitative safety and topological continuity and provide an alternative characterization of quantitative safety and liveness in terms of their boolean counterparts. For all common value functions, we show how the safety closure of a quantitative automaton can be constructed in PTime, and we provide PSpace-complete checks of whether a given quantitative automaton is safe or live, with the exception of LimInfAvg and LimSupAvg automata, for which the safety check is in ExpSpace. Moreover, for deterministic Sup, LimInf, and LimSup automata, we give PTime decompositions into safe and live automata. These decompositions enable the separation of techniques for safety and liveness verification for quantitative specifications.},
  author       = {Boker, Udi and Henzinger, Thomas A and Mazzocchi, Nicolas Adrien and Sarac, Naci E},
  booktitle    = {34th International Conference on Concurrency Theory},
  isbn         = {9783959772990},
  issn         = {1868-8969},
  location     = {Antwerp, Belgium},
  publisher    = {Schloss Dagstuhl - Leibniz-Zentrum für Informatik},
  title        = {{Safety and liveness of quantitative automata}},
  doi          = {10.4230/LIPIcs.CONCUR.2023.17},
  volume       = {279},
  year         = {2023},
}

@article{14466,
  abstract     = {The first long-lived turbulent structures observable in planar shear flows take the form of localized stripes, inclined with respect to the mean flow direction. The dynamics of these stripes is central to transition, and recent studies proposed an analogy to directed percolation where the stripes’ proliferation is ultimately responsible for the turbulence becoming sustained. In the present study we focus on the internal stripe dynamics as well as on the eventual stripe expansion, and we compare the underlying mechanisms in pressure- and shear-driven planar flows, respectively, plane-Poiseuille and plane-Couette flow. Despite the similarities of the overall laminar–turbulence patterns, the stripe proliferation processes in the two cases are fundamentally different. Starting from the growth and sustenance of individual stripes, we find that in plane-Couette flow new streaks are created stochastically throughout the stripe whereas in plane-Poiseuille flow streak creation is deterministic and occurs locally at the downstream tip. Because of the up/downstream symmetry, Couette stripes, in contrast to Poiseuille stripes, have two weak and two strong laminar turbulent interfaces. These differences in symmetry as well as in internal growth give rise to two fundamentally different stripe splitting mechanisms. In plane-Poiseuille flow splitting is connected to the elongational growth of the original stripe, and it results from a break-off/shedding of the stripe's tail. In plane-Couette flow splitting follows from a broadening of the original stripe and a division along the stripe into two slimmer stripes.},
  author       = {Marensi, Elena and Yalniz, Gökhan and Hof, Björn},
  issn         = {1469-7645},
  journal      = {Journal of Fluid Mechanics},
  keywords     = {turbulence, transition to turbulence, patterns},
  publisher    = {Cambridge University Press},
  title        = {{Dynamics and proliferation of turbulent stripes in plane-Poiseuille and plane-Couette flows}},
  doi          = {10.1017/jfm.2023.780},
  volume       = {974},
  year         = {2023},
}

@article{13274,
  abstract     = {Viscous flows through pipes and channels are steady and ordered until, with increasing velocity, the laminar motion catastrophically breaks down and gives way to turbulence. How this apparently discontinuous change from low- to high-dimensional motion can be rationalized within the framework of the Navier-Stokes equations is not well understood. Exploiting geometrical properties of transitional channel flow we trace turbulence to far lower Reynolds numbers (Re) than previously possible and identify the complete path that reversibly links fully turbulent motion to an invariant solution. This precursor of turbulence destabilizes rapidly with Re, and the accompanying explosive increase in attractor dimension effectively marks the transition between deterministic and de facto stochastic dynamics.},
  author       = {Paranjape, Chaitanya S and Yalniz, Gökhan and Duguet, Yohann and Budanur, Nazmi B and Hof, Björn},
  issn         = {1079-7114},
  journal      = {Physical Review Letters},
  keywords     = {General Physics and Astronomy},
  number       = {3},
  publisher    = {American Physical Society},
  title        = {{Direct path from turbulence to time-periodic solutions}},
  doi          = {10.1103/physrevlett.131.034002},
  volume       = {131},
  year         = {2023},
}

@inproceedings{13310,
  abstract     = {Machine-learned systems are in widespread use for making decisions about humans, and it is important that they are fair, i.e., not biased against individuals based on sensitive attributes. We present runtime verification of algorithmic fairness for systems whose models are unknown, but are assumed to have a Markov chain structure. We introduce a specification language that can model many common algorithmic fairness properties, such as demographic parity, equal opportunity, and social burden. We build monitors that observe a long sequence of events as generated by a given system, and output, after each observation, a quantitative estimate of how fair or biased the system was on that run until that point in time. The estimate is proven to be correct modulo a variable error bound and a given confidence level, where the error bound gets tighter as the observed sequence gets longer. Our monitors are of two types, and use, respectively, frequentist and Bayesian statistical inference techniques. While the frequentist monitors compute estimates that are objectively correct with respect to the ground truth, the Bayesian monitors compute estimates that are correct subject to a given prior belief about the system’s model. Using a prototype implementation, we show how we can monitor if a bank is fair in giving loans to applicants from different social backgrounds, and if a college is fair in admitting students while maintaining a reasonable financial burden on the society. Although they exhibit different theoretical complexities in certain cases, in our experiments, both frequentist and Bayesian monitors took less than a millisecond to update their verdicts after each observation.},
  author       = {Henzinger, Thomas A and Karimi, Mahyar and Kueffner, Konstantin and Mallik, Kaushik},
  booktitle    = {Computer Aided Verification},
  isbn         = {9783031377020},
  issn         = {1611-3349},
  location     = {Paris, France},
  pages        = {358–382},
  publisher    = {Springer Nature},
  title        = {{Monitoring algorithmic fairness}},
  doi          = {10.1007/978-3-031-37703-7_17},
  volume       = {13965},
  year         = {2023},
}

@inproceedings{12859,
  abstract     = {Machine learning models are vulnerable to adversarial perturbations, and a thought-provoking paper by Bubeck and Sellke has analyzed this phenomenon through the lens of over-parameterization: interpolating smoothly the data requires significantly more parameters than simply memorizing it. However, this "universal" law provides only a necessary condition for robustness, and it is unable to discriminate between models. In this paper, we address these gaps by focusing on empirical risk minimization in two prototypical settings, namely, random features and the neural tangent kernel (NTK). We prove that, for random features, the model is not robust for any degree of over-parameterization, even when the necessary condition coming from the universal law of robustness is satisfied. In contrast, for even activations, the NTK model meets the universal lower bound, and it is robust as soon as the necessary condition on over-parameterization is fulfilled. This also addresses a conjecture in prior work by Bubeck, Li and Nagaraj. Our analysis decouples the effect of the kernel of the model from an "interaction matrix", which describes the interaction with the test data and captures the effect of the activation. Our theoretical results are corroborated by numerical evidence on both synthetic and standard datasets (MNIST, CIFAR-10).},
  author       = {Bombari, Simone and Kiyani, Shayan and Mondelli, Marco},
  booktitle    = {Proceedings of the 40th International Conference on Machine Learning},
  location     = {Honolulu, HI, United States},
  pages        = {2738--2776},
  publisher    = {ML Research Press},
  title        = {{Beyond the universal law of robustness: Sharper laws for random features and neural tangent kernels}},
  volume       = {202},
  year         = {2023},
}

@phdthesis{12491,
  abstract     = {The extracellular matrix (ECM) is a hydrated and complex three-dimensional network consisting of proteins, polysaccharides, and water. It provides structural scaffolding for the cells embedded within it and is essential in regulating numerous physiological processes, including cell migration and proliferation, wound healing, and stem cell fate. 
Despite extensive study, detailed structural knowledge of ECM components in physiologically relevant conditions is still rudimentary. This is due to methodological limitations in specimen preparation protocols which are incompatible with keeping large samples, such as the ECM, in their native state for subsequent imaging. Conventional electron microscopy (EM) techniques rely on fixation, dehydration, contrasting, and sectioning. This results in the alteration of a highly hydrated environment and the potential introduction of artifacts. Other structural biology techniques, such as nuclear magnetic resonance (NMR) spectroscopy and X-ray crystallography, allow high-resolution analysis of protein structures but only work on homogenous and purified samples, hence lacking contextual information. Currently, no approach exists for the ultrastructural and structural study of extracellular components under native conditions in a physiological, 3D environment. 
In this thesis, I have developed a workflow that allows for the ultrastructural analysis of the ECM in near-native conditions at molecular resolution. The developments I introduced include implementing a novel specimen preparation workflow for cell-derived matrices (CDMs) to render them compatible with ion-beam milling and subsequent high-resolution cryo-electron tomography (ET). 
To this end, I have established protocols to generate CDMs grown over several weeks on EM grids that are compatible with downstream cryo-EM sample preparation and imaging techniques. Characterization of these ECMs confirmed that they contain essential ECM components such as collagen I, collagen VI, and fibronectin I in high abundance and hence represent a bona fide biologically-relevant sample. I successfully optimized vitrification of these specimens by testing various vitrification techniques and cryoprotectants. 
In order to obtain high-resolution molecular insights into the ultrastructure and organization of CDMs, I established cryo-focused ion beam scanning electron microscopy (FIBSEM) on these challenging and complex specimens. I explored different approaches for the creation of thin cryo-lamellae by FIB milling and succeeded in optimizing the cryo-lift-out technique, resulting in high-quality lamellae of approximately 200 nm thickness. 
High-resolution Cryo-ET of these lamellae revealed for the first time the architecture of native CDM in the context of matrix-secreting cells. This allowed for the in situ visualization of fibrillar matrix proteins such as collagen, laying the foundation for future structural and ultrastructural characterization of these proteins in their near-native environment. 
In summary, in this thesis, I present a novel workflow that combines state-of-the-art cryo-EM specimen preparation and imaging technologies to permit characterization of the ECM, an important tissue component in higher organisms. This innovative and highly versatile workflow will enable addressing far-reaching questions on ECM architecture, composition, and reciprocal ECM-cell interactions.},
  author       = {Zens, Bettina},
  isbn         = {978-3-99078-027-5},
  issn         = {2663-337X},
  keywords     = {cryo-EM, cryo-ET, FIB milling, method development, FIBSEM, extracellular matrix, ECM, cell-derived matrices, CDMs, cell culture, high pressure freezing, HPF, structural biology, tomography, collagen},
  pages        = {187},
  publisher    = {Institute of Science and Technology Austria},
  title        = {{Ultrastructural characterization of natively preserved extracellular matrix by cryo-electron tomography}},
  doi          = {10.15479/at:ista:12491},
  year         = {2023},
}

