Access to raw network traffic data is essential for many computer networking tasks, from traffic modeling to performance evaluation. Unfortunately, this data is scarce due to high collection costs and governance rules. Previous efforts explore this challenge by generating synthetic network data, but fail to reliably handle multi-flow sessions, struggle to reason about stateful communication in moderate to long-duration network sessions, and lack robust evaluations tied to real-world utility. We propose a new method based on state-space models called NetSSM that generates raw network traffic at the packet-level granularity. Our approach captures interactions between multiple, interleaved flows -- an objective unexplored in prior work -- and effectively reasons about flow-state in sessions to capture traffic characteristics. NetSSM accomplishes this by learning from and producing traces 8x and 78x longer than existing transformer-based approaches. Evaluation results show that our method generates high-fidelity traces that outperform prior efforts in existing benchmarks. We also find that NetSSM's traces have high semantic similarity to real network data regarding compliance with standard protocol requirements and flow and session-level traffic characteristics.

Machine learning has shown tremendous potential for improving the capabilities of network traffic analysis applications, often outperforming simpler rule-based heuristics. However, ML-based solutions remain difficult to deploy in practice. Many existing approaches only optimize the predictive performance of their models, overlooking the practical challenges of running them against network traffic in real time. This is especially problematic in the domain of traffic analysis, where the efficiency of the serving pipeline is a critical factor in determining the usability of a model. In this work, we introduce CATO, a framework that addresses this problem by jointly optimizing the predictive performance and the associated systems costs of the serving pipeline. CATO leverages recent advances in multi-objective Bayesian optimization to efficiently identify Pareto-optimal configurations, and automatically compiles end-to-end optimized serving pipelines that can be deployed in real networks. Our evaluations show that compared to popular feature optimization techniques, CATO can provide up to 3600x lower inference latency and 3.7x higher zero-loss throughput while simultaneously achieving better model performance.

The concept of linearity plays a central role in both mathematics and computer science, with distinct yet complementary meanings. In mathematics, linearity underpins functions and vector spaces, forming the foundation of linear algebra and functional analysis. In computer science, it relates to resource-sensitive computation. Linear Logic (LL), for instance, models assumptions that must be used exactly once, providing a natural framework for tracking computational resources such as time, memory, or data access. This dual perspective makes linearity essential to programming languages, type systems, and formal models that express both computational complexity and composability. Bridging these interpretations enables rigorous yet practical methodologies for analyzing and verifying complex systems. This thesis explores the use of LL to model programming paradigms based on linearity. It comprises two parts: ADLL and CryptoBLL. The former applies LL to Automatic Differentiation (AD), modeling linear functions over the reals and the transposition operation. The latter uses LL to express complexity constraints on adversaries in computational cryptography. In AD, two main approaches use linear type systems: a theoretical one grounded in proof theory, and a practical one implemented in JAX, a Python library developed by Google for machine learning research. In contrast, frameworks like PyTorch and TensorFlow support AD without linear types. ADLL aims to bridge theory and practice by connecting JAX's type system to LL. In modern cryptography, several calculi aim to model cryptographic proofs within the computational paradigm. These efforts face a trade-off between expressiveness, to capture reductions, and simplicity, to abstract probability and complexity. CryptoBLL addresses this tension by proposing a framework for the automatic analysis of protocols in computational cryptography.

The Brownian sphere is a random metric space, homeomorphic to the two-dimensional sphere, which arises as the universal scaling limit of many types of random planar maps. The direct construction of the Brownian sphere is via a continuous analogue of the Cori--Vauquelin--Schaeffer (CVS) bijection. The CVS bijection maps labeled trees to planar maps, and the continuous version maps Aldous' continuum random tree with Brownian labels (the Brownian snake) to the Brownian sphere. In this work, we describe the inverse of the continuous CVS bijection, by constructing the Brownian snake as a measurable function of the Brownian sphere. Special care is needed to work with the orientation of the Brownian sphere.

Jellyfish galaxies provide direct evidence of ram pressure stripping in cluster environments. We investigate the role of magnetic fields in the formation of jellyfish galaxies with a multiphase interstellar medium (ISM) using radiation magneto-hydrodynamic simulations. We impose magnetized (MHD) and non-magnetized (HD) winds on the gas-rich dwarf galaxies containing the magnetized or non-magnetized ISM. The MHD winds strip the disk gas more effectively than the HD winds because of the magnetic force acting against the local density gradient, which results in remarkably different ram pressure stripped features. The magnetic fields induced by the MHD winds generate a strong magnetic pressure, which forms smoothed disks and tail gas features. Since the stripped ISM in MHD wind cases travels while being nearly isolated from the intracluster medium (ICM), the stripped ISM mostly forms stars within 20~kpc of the galactic disks. In contrast, non-magnetized winds facilitate the efficient mixing of the stripped ISM with the ICM, resulting in the formation of abundant warm clouds that cool and collapse in the distant ($\sim50-100\,$kpc) tails at times of a few hundred Myr. Consequently, distant tail star formation occurs only in the HD wind runs. Finally, despite the different tail features, the star formation rates in the disk remain similar owing to the interplay between the increased gas stripping and the gas density increase in the disks of the MHD wind runs. These results suggest that the magnetized ICM may have a significant influence on jellyfish galaxies, whereas the magnetized ISM play a minor role.

As Machine Learning (ML) is still a recent field of study, especially outside the realm of abstract Mathematics and Computer Science, few works have been conducted on the political aspect of large Language Models (LLMs), and more particularly about the alignment process and its political dimension. This process can be as simple as prompt engineering but is also very complex and can affect completely unrelated notions. For example, politically directed alignment has a very strong impact on an LLM's embedding space and the relative position of political notions in such a space. Using special tools to evaluate general political bias and analyze the effects of alignment, we can gather new data to understand its causes and possible consequences on society. Indeed, by taking a socio-political approach, we can hypothesize that most big LLMs are aligned with what Marxist philosophy calls the 'dominant ideology.' As AI's role in political decision-making, at the citizen's scale but also in government agencies, such biases can have huge effects on societal change, either by creating new and insidious pathways for societal uniformity or by allowing disguised extremist views to gain traction among the people.

Mathematicians have been proposing for sometimes that Monge-Ampère equation, a nonlinear generalization of the Poisson equation, where trace of the Hessian is replaced by its determinant, provides an alternative non-relativistic description of gravity. Monge-Ampère equation is affine invariant, has rich geometric properties, connects to optimal transport theory, and remains bounded at short distances. Monge-Ampère gravity, that uses a slightly different form of the Monge-Ampère equation, naturally emerges through the application of large-deviation principle to a Brownian system of indistinguishable and independent particles. In this work we provide a physical formulation of this mathematical model, study its theoretical viability and confront it with observations. We show that Monge-Ampère gravity cannot replace the Newtonian gravity as it does not withstand the solar-system test. We then show that Monge-Ampère gravity can describe a scalar field, often evoked in modified theories of gravity such as Galileons. We show that Monge-Ampère gravity, as a nonlinear model of a new scalar field, is screened at short distances, and behaves differently from Newtonian gravity above galactic scales but approaches it asymptotically. Finally, we write a relativistic Lagrangian for Monge-Ampère gravity in flat space time, which is the field equation of a sum of the Lagrangians of all Galileons. We also show how the Monge-Ampère equation can be obtained from the fully covariant Lagrangian of quartic Galileon in the static limit. The connection between optimal transport theory and modified theories of gravity with second-order field equations, unravelled here, remains a promising domain to further explore.

We present an analytical determination of the star formation rate (SFR) in molecular clouds, based on a time-dependent extension of our analytical theory of the stellar initial mass function (IMF). The theory yields SFR's in good agreement with observations, suggesting that turbulence {\it is} the dominant, initial process responsible for star formation. In contrast to previous SFR theories, the present one does not invoke an ad-hoc density threshold for star formation; instead, the SFR {\it continuously} increases with gas density, naturally yielding two different characteristic regimes, thus two different slopes in the SFR vs gas density relationship, in agreement with observational determinations. Besides the complete SFR derivation, we also provide a simplified expression, which reproduces reasonably well the complete calculations and can easily be used for quick determinations of SFR's in cloud environments. A key property at the heart of both our complete and simplified theory is that the SFR involves a {\it density-dependent dynamical time}, characteristic of each collapsing (prestellar) overdense region in the cloud, instead of one single mean or critical freefall timescale. Unfortunately, the SFR also depends on some ill determined parameters, such as the core-to-star mass conversion efficiency and the crossing timescale. Although we provide estimates for these parameters, their uncertainty hampers a precise quantitative determination of the SFR, within less than a factor of a few.

We present a study of the ion stopping power due to free and bound electrons in a warm dense plasma. Our main goal is to propose a method of stopping-power calculation expected to be valid for any ionization degree. The free-electron contribution is described by the Maynard-Deutsch-Zimmerman formula and the bound-electron one relies on the Bethe formula with corrections, in particular taking into account density and shell effects. The impact of the bound-state computation by three different parametric potentials is investigated within the Garbet formalism for the mean excitation energy. The first parametric potential is due to Green, Sellin and Zachor, the second one was proposed by Yunta, and the third one was introduced by Klapisch in the framework of atomic-structure computations. The results are compared to the ones of self-consistent average-atom calculations. This approach correctly bridges the limits of neutral and fully-ionized matter.

A "2-group" is a category equipped with a multiplication satisfying laws like those of a group. Just as groups have representations on vector spaces, 2-groups have representations on "2-vector spaces", which are categories analogous to vector spaces. Unfortunately, Lie 2-groups typically have few representations on the finite-dimensional 2-vector spaces introduced by Kapranov and Voevodsky. For this reason, Crane, Sheppeard and Yetter introduced certain infinite-dimensional 2-vector spaces called "measurable categories" (since they are closely related to measurable fields of Hilbert spaces), and used these to study infinite-dimensional representations of certain Lie 2-groups. Here we continue this work. We begin with a detailed study of measurable categories. Then we give a geometrical description of the measurable representations, intertwiners and 2-intertwiners for any skeletal measurable 2-group. We study tensor products and direct sums for representations, and various concepts of subrepresentation. We describe direct sums of intertwiners, and sub-intertwiners - features not seen in ordinary group representation theory. We study irreducible and indecomposable representations and intertwiners. We also study "irretractable" representations - another feature not seen in ordinary group representation theory. Finally, we argue that measurable categories equipped with some extra structure deserve to be considered "separable 2-Hilbert spaces", and compare this idea to a tentative definition of 2-Hilbert spaces as representation categories of commutative von Neumann algebras.

The growth of galaxies in the early Universe is driven by accretion of circum- and inter-galactic gas. Simulations predict that steady streams of cold gas penetrate the dark matter halos of galaxies, providing the raw material necessary to sustain star formation. We report a filamentary stream of gas that extends for 100 kiloparsecs and connects to the massive radio galaxy 4C 41.17. The stream is detected using sub-millimeter observations of the [CI] line of atomic carbon, a tracer of neutral atomic or molecular hydrogen gas. The galaxy contains a central gas reservoir that is fueling a vigorous starburst. Our results show that the raw material for star formation can be present in cosmic streams outside galaxies.

Now that large databases of resolved galaxy images are provided by modern imaging surveys, advanced morphological studies can be envisioned, urging for well defined calibration samples. We present the EFIGI catalogue, a multiwavelength database specifically designed for a dense sampling of all Hubble types. The catalogue merges data from standard surveys and catalogues (Principal Galaxy Catalogue, Sloan Digital Sky Survey, Value-Added Galaxy Catalogue, HyperLeda, and the NASA Extragalactic Database) and provides detailed morphological information. Imaging data are obtained from the SDSS DR4 in the u, g, r, i, and z bands for a sample of 4458 PGC galaxies, whereas photometric and spectroscopic data are obtained from the SDSS DR5 catalogue. Point-Spread Function models are derived in all five bands. Composite colour images of all objects are visually examined by a group of astronomers, and galaxies are staged along the Hubble sequence and classified according to 16 morphological attributes describing their structure, texture, as well as environment and appearance on a five-level scale. The EFIGI Hubble sequence shows remarkable agreement with the RC3 Revised Hubble Sequence. The main characteristics and reliability of the catalogue are examined, including photometric completeness, type mix, systematic trends and correlations. The final EFIGI database is a large sub-sample of the local Universe, with a dense sampling of Sd, Sdm, Sm and Im types compared to magnitude-limited catalogues. We estimate the photometric catalogue to be more than ~ 80% complete for galaxies with 10

Graph embedding approaches attempt to project graphs into geometric entities, i.e, manifolds. The idea is that the geometric properties of the projected manifolds are helpful in the inference of graph properties. However, if the choice of the embedding manifold is incorrectly performed, it can lead to incorrect geometric inference. In this paper, we argue that the classical embedding techniques cannot lead to correct geometric interpretation as they miss the curvature at each point, of manifold. We advocate that for doing correct geometric interpretation the embedding of graph should be done over regular constant curvature manifolds. To this end, we present an embedding approach, the discrete Ricci flow graph embedding (dRfge) based on the discrete Ricci flow that adapts the distance between nodes in a graph so that the graph can be embedded onto a constant curvature manifold that is homogeneous and isotropic, i.e., all directions are equivalent and distances comparable, resulting in correct geometric interpretations. A major contribution of this paper is that for the first time, we prove the convergence of discrete Ricci flow to a constant curvature and stable distance metrics over the edges. A drawback of using the discrete Ricci flow is the high computational complexity that prevented its usage in large-scale graph analysis. Another contribution of this paper is a new algorithmic solution that makes it feasible to calculate the Ricci flow for graphs of up to 50k nodes, and beyond. The intuitions behind the discrete Ricci flow make it possible to obtain new insights into the structure of large-scale graphs. We demonstrate this through a case study on analyzing the internet connectivity structure between countries at the BGP level.

It is reported that the numerical simulations of the Fermi-Pasta-Ulam problem were performed by a young lady, Mary Tsingou. After 50 years of omission, it is time for a proper recognition of her decisive contribution to the first ever numerical experiment, central in the solitons and chaos theories, but also one of the very first out-of-equilibrium statistical mechanics study. Let us quote from now on the Fermi-Pasta-Ulam-Tsingou problem.

In the last decade, the photospheric solar metallicity as determined from spectroscopy experienced a remarkable downward revision. Part of this effect can be attributed to an improvement of atomic data and the inclusion of NLTE computations, but also the use of hydrodynamical model atmospheres seemed to play a role. This "decrease" with time of the metallicity of the solar photosphere increased the disagreement with the results from helioseismology. With a CO5BOLD 3D model of the solar atmosphere, the CIFIST team at the Paris Observatory re-determined the photospheric solar abundances of several elements, among them C, N, and O. The spectroscopic abundances are obtained by fitting the equivalent width and/or the profile of observed spectral lines with synthetic spectra computed from the 3D model atmosphere. We conclude that the effects of granular fluctuations depend on the characteristics of the individual lines, but are found to be relevant only in a few particular cases. 3D effects are not reponsible for the systematic lowering of the solar abundances in recent years. The solar metallicity resulting from this analysis is Z=0.0153, Z/X=0.0209.

Accurate and efficient network traffic classification is important for many network management tasks, from traffic prioritization to anomaly detection. Although classifiers using pre-computed flow statistics (e.g., packet sizes, inter-arrival times) can be efficient, they may experience lower accuracy than techniques based on raw traffic, including packet captures. Past work on representation learning-based classifiers applied to network traffic captures has shown to be more accurate, but slower and requiring considerable additional memory resources, due to the substantial costs in feature preprocessing. In this paper, we explore this trade-off and develop the Adaptive Constraint-Driven Classification (AC-DC) framework to efficiently curate a pool of classifiers with different target requirements, aiming to provide comparable classification performance to complex packet-capture classifiers while adapting to varying network traffic load. AC-DC uses an adaptive scheduler that tracks current system memory availability and incoming traffic rates to determine the optimal classifier and batch size to maximize classification performance given memory and processing constraints. Our evaluation shows that AC-DC improves classification performance by more than 100% compared to classifiers that rely on flow statistics alone; compared to the state-of-the-art packet-capture classifiers, AC-DC achieves comparable performance (less than 12.3% lower in F1-Score), but processes traffic over 150x faster.

The formation mechanism of Brown Dwarfs (BDs), whether akin to stars or ejected planetary-mass objects, remains debated. We present the first 3D radiation-MHD simulations of magnetized, turbulent, gravitationally unstable low-mass cores ($0.05-0.1\ \mathrm{M_{\odot}}$) collapsing into proto-BDs. Using the {\ttfamily RAMSES} code with adaptive mesh refinement, we model the full dynamical range ($10^{5}~-10^{22}\ \mathrm{cm^{-3}}$), including radiative transfer (flux limited diffusion) and non-ideal MHD (ambipolar diffusion). Our simulations self-consistently follow the isothermal collapse, first hydrostatic core formation, H$_{2}$ dissociation, and BD birth. The resulting BDs have initial radii $\approx 0.75\ \mathrm{R_{\odot}}$ and masses $\approx 0.8\ \mathrm{M_{Jup}}$, growing via accretion as we follow the early evolution of the object. Crucially, we find that BDs may form similarly to low-mass stars but with a prolonged first-core phase, supporting a star-like formation scenario.

We consider two-parametric families of non-autonomous ordinary differential equations on the two-torus with the coordinates $(x,t)$ of the type $\dot x=v(x)+A+Bf(t)$. We study its rotation number as a function of the parameters $(A,B)$. The {\it phase-lock areas} are those level sets of the rotation number function $\rho=\rho(A,B)$ that have non-empty interiors. this http URL, this http URL, this http URL have studied the case, when $v(x)=\sin x$ in their joint paper. They have observed the quantization effect: for every smooth periodic function $f(t)$ the family of equations may have phase-lock areas only for integer rotation numbers. Another proof of this quantization statement was later obtained in a joint paper by this http URL, this http URL, this http URL. This implies the similar quantization effect for every $v(x)=a\sin(mx)+b\cos(mx)+c$ and rotation numbers that are multiples of $\frac 1m$. We show that for every other analytic vector field $v(x)$ (i.e., having at least two Fourier harmonics with non-zero non-opposite degrees and nonzero coefficients) there exists an analytic periodic function $f(t)$ such that the corresponding family of equations has phase-lock areas for all the rational values of the rotation number.

We present a comprehensive review of the physical behavior of yield stress materials in soft condensed matter, which encompass a broad range of materials from colloidal assemblies and gels to emulsions and non-Brownian suspensions. All these disordered materials display a nonlinear flow behavior in response to external mechanical forces, due to the existence of a finite force threshold for flow to occur: the yield stress. We discuss both the physical origin and rheological consequences associated with this nonlinear behavior, and give an overview of experimental techniques available to measure the yield stress. We discuss recent progress concerning a microscopic theoretical description of the flow dynamics of yield stress materials, emphasizing in particular the role played by relaxation time scales, the interplay between shear flow and aging behavior, the existence of inhomogeneous shear flows and shear bands, wall slip, and non-local effects in confined geometries.

Loop quantum cosmology tries to capture the main ideas of loop quantum gravity and to apply them to the Universe as a whole. Two main approaches within this framework have been considered to date for the study of cosmological perturbations: the dressed metric approach and the deformed algebra approach. They both have advantages and drawbacks. In this article, we accurately compare their predictions. In particular, we compute the associated primordial tensor power spectra. We show -- numerically and analytically -- that the large scale behavior is similar for both approaches and compatible with the usual prediction of general relativity. The small scale behavior is, the other way round, drastically different. Most importantly, we show that in a range of wavenumbers explicitly calculated, both approaches do agree on predictions that, in addition, differ from standard general relativity and do not depend on unknown parameters. These features of the power spectrum at intermediate scales might constitute a universal loop quantum cosmology prediction that can hopefully lead to observational tests and constraints. We also present a complete analytical study of the background evolution for the bouncing universe that can be used for other purposes.