Search

We begin the systematic study of cohomological Hecke operators of modifications of coherent sheaves on a smooth surface $X$, along a fixed proper curve $Z \subset X$. We develop the necessary geometric foundations in order to define the $T$-equivariant cohomological Hall algebra $\mathbf{HA}^{\mathbf{D}, T}_{X,Z}$ of the moduli stack of coherent sheaves on $X$ with set-theoretic support on $Z$, in the setting of a general motivic formalism $\mathbf{D}$. The algebra $\mathbf{HA}^{\mathbf{D}, A}_{X,Z}$ is functorial with respect to closed immersions $Z' \subset Z$ and transformations of the motivic formalism $\mathbf{D}$, and only depends on the formal neighborhood $\widehat{X}_Z$ of $Z$ in $X$. Assume $\mathbf{D}$ gives rise to Borel-Moore homology. When $X$ is a resolution of a Kleinian singularity and $Z$ is the exceptional divisor, we explicitly identify $\mathbf{HA}^T_{X,Z}$ with a completed nonstandard positive half $\mathbb{Y}^+_\infty(\mathfrak{g})$ of the affine Yangian $\mathbb{Y}(\mathfrak{g})$ of the corresponding affine ADE type Lie algebra $\mathfrak{g}$. Let $Z_1,Z_2 \subset X$ be curves with $Z_1\cap Z_2$ being zero-dimensional. We conjecture a PBW type theorem relating the COHAs of the pairs $(X, Z_1\cup Z_2), (X, Z_1)$, and $(X, Z_2)$ and partially prove it in several interesting cases, including all Kleinian resolutions of singularities and elliptic surfaces of types $D$ and $E$. Our main tools are: (i) a continuity theorem describing the behavior of COHAs of objects in the heart of $t$-structures $\tau_n$ when the sequence $(\tau_n)_n$ converges in an appropriate sense to a fixed $t$-structure $\tau_\infty$; (ii) a theorem relating the action of the braid group $B_Q$ by derived autoequivalences on the preprojective algebra of a quiver $Q$ with the algebraic action of $B_Q$ on the associated COHA $\mathbf{HA}^T_Q$., Comment: v1: 335 pages

Analog in-memory computing is an emerging paradigm designed to efficiently accelerate deep neural network workloads. Recent advancements have focused on either inference or training acceleration. However, a unified analog in-memory technology platform-capable of on-chip training, weight retention, and long-term inference acceleration-has yet to be reported. This work presents an all-in-one analog AI accelerator, combining these capabilities to enable energy-efficient, continuously adaptable AI systems. The platform leverages an array of analog filamentary conductive-metal-oxide (CMO)/HfOx resistive switching memory cells (ReRAM) integrated into the back-end-of-line (BEOL). The array demonstrates reliable resistive switching with voltage amplitudes below 1.5 V, compatible with advanced technology nodes. The array's multi-bit capability (over 32 stable states) and low programming noise (down to 10 nS) enable a nearly ideal weight transfer process, more than an order of magnitude better than other memristive technologies. Inference performance is validated through matrix-vector multiplication simulations on a 64x64 array, achieving a root-mean-square error improvement by a factor of 10 at 1 second and 3 at 10 years after programming, compared to state-of-the-art. Training accuracy closely matching the software equivalent is achieved across different datasets. The CMO/HfOx ReRAM technology lays the foundation for efficient analog systems accelerating both inference and training in deep neural networks.

Objective: This work introduces a framework for multivariate time series analysis aimed at detecting and quantifying collective emerging behaviors in the dynamics of physiological networks. Methods: Given a network system mapped by a vector random process, we compute the predictive information (PI) between the present and past network states and dissect it into amounts quantifying the unique, redundant and synergistic information shared by the present of the network and the past of each unit. Emergence is then quantified as the prevalence of the synergistic over the redundant contribution. The framework is implemented in practice using vector autoregressive (VAR) models. Results: Validation in simulated VAR processes documents that emerging behaviors arise in networks where multiple causal interactions coexist with internal dynamics. The application to cardiovascular and respiratory networks mapping the beat-to-beat variability of heart rate, arterial pressure and respiration measured at rest and during postural stress reveals the presence of statistically significant net synergy, as well as its modulation with sympathetic nervous system activation. Conclusion: Causal emergence can be efficiently assessed decomposing the PI of network systems via VAR models applied to multivariate time series. This approach evidences the synergy/redundancy balance as a hallmark of integrated short-term autonomic control in cardiovascular and respiratory networks. Significance: Measures of causal emergence provide a practical tool to quantify the mechanisms of causal influence that determine the dynamic state of cardiovascular and neural network systems across distinct physiopathological conditions.

We consider the two-dimensional $\mathbb{Z}_{2}$ Ising gauge theory coupled to fermionic matter. In absence of electric fields, we prove that, at half-filling, the ground state of the gauge theory coincides with the $\pi$-flux phase, associated with magnetic flux equal to $\pi$ in every elementary lattice plaquette, provided the fermionic hopping is large enough. This proves in particular the semimetallic behavior of the ground state of the model. Furthermore, we compute the magnetic susceptibility of the gauge theory, and we prove that it is given by the one of massless $2d$ Dirac fermions, thus rigorously justifying recent numerical computations. The proof is based on reflection positivity and chessboard estimates, and on lattice conservation laws for the computation of the transport coefficient., Comment: 61 pages, 11 figures

The vertical plane transverse emittance of accelerated electron bunches at the AWAKE experiment at CERN has been determined, using three different methods of data analysis. This is a proof-of-principle measurement using the existing AWAKE electron spectrometer to validate the measurement technique. Large values of the geometric emittance, compared to that of the injection beam, are observed ($\sim \SI{0.5}{\milli\metre\milli\radian}$ compared with $\sim \SI{0.08}{\milli\metre\milli\radian}$), which is in line with expectations of emittance growth arising from plasma density ramps and large injection beam bunch size. Future iterations of AWAKE are anticipated to operate in conditions where emittance growth is better controlled, and the effects of the imaging systems of the existing and future spectrometer designs on the ability to measure the emittance are discussed. Good performance of the instrument down to geometric emittances of approximately $\SI{1e-4}{\milli\metre\milli\radian}$ is required, which may be possible with improved electron optics and imaging., Comment: 20 pages, 9 figures

We consider the transport properties of non-interacting, gapless one-dimensional quantum systems and of the edge modes of two-dimensional topological insulators, in the presence of time-dependent perturbations. We prove the validity of Kubo formula, in the zero temperature and infinite volume limit, for a class of perturbations that are weak and slowly varying in space and in time, in an Euler-like scaling. The proof relies on the representation of the real time Duhamel series in imaginary time, which allows to prove its convergence uniformly in the scaling parameter and in the size of system, at low temperatures. Furthermore, it allows to exploit a suitable cancellation for the scaling limit of the model, related to the emergent anomalous chiral gauge symmetry of relativistic one-dimensional fermions. The cancellation implies that, as the temperature and the scaling parameter are sent to zero, the linear response is the only contribution to the full response of the system. The explicit form of the leading contribution to the response function is determined by lattice conservation laws. In particular, the method allows to prove the quantization of the edge conductance of $2d$ quantum Hall systems from quantum dynamics., Comment: 56 pages, 3 figures

Robotic manipulation of fresh fruits and vegetables, including the grasping of multiple loose items, has a strong industrial need but it still is a challenging task for robotic manipulation. This paper outlines the distinctive manipulation strategies used by humans to pick loose fruits and vegetables with the aim to better adopt them for robotic manipulation of diverse items. In this work we present a first version of a robotic setup designed to pick different single or multiple fresh items, featuring multi-fingered compliant robotic gripper. We analyse human grasping strategies from the perspective of industrial Key Performance Indicators (KPIs) used in the logistic sector. The robotic system was validated using the same KPIs, as well as taking into account human performance and strategies. This paper lays the foundation for future development of the robotic demonstrator for fresh fruit and vegetable intelligent manipulation, and outlines the need for generic approaches to handle the complexity of the task., Comment: *Authors contributed equally

In this paper we establish combinatorial characterisations of symmetry-generic infinitesimally rigid frameworks in the Euclidean plane for rotational groups of order 4 and 6, and of odd order between 5 and 1000, where a joint may lie at the centre of rotation. This extends the corresponding results for these groups in the free action case obtained by R. Ikeshita and S. Tanigawa in 2015, and our recent results for the reflection group and the rotational groups of order 2 and 3 in the non-free action case. The characterisations are given in terms of sparsity counts on the corresponding group-labelled quotient graphs, and are obtained via symmetry-adapted versions of recursive Henneberg-type graph constructions. For rotational groups of even order at least 8, we show that the sparsity counts alone are not sufficient for symmetry-generic infinitesimal rigidity., Comment: 25 pages (excluding references), 7 figures

We consider Haldane-like $2d$ topological insulators on the cylinder, in the presence of weak quasi-periodic disorder. We prove that, at large distances, the boundary correlations agree with the correlations of a renormalized, translation-invariant, massless relativistic model in $1+1$ dimensions, multiplied by non-universal oscillatory factors, incommensurate with the lattice spacing. Furthermore, we compute the edge conductance and the edge susceptibility, starting from Kubo formula. We obtain explicit expressions for these response functions, completely determined by the renormalized Fermi velocity of the edge modes. In particular, we prove the quantization of the edge conductance, and the non-universality of the susceptibility. The proof relies on multiscale analysis and rigorous renormalization group methods for quasi-periodic systems, and on lattice Ward identities., Comment: 92 pages, 5 figures

In recent years, drone detection has quickly become a subject of extreme interest: the potential for fast-moving objects of contained dimensions to be used for malicious intents or even terrorist attacks has posed attention to the necessity for precise and resilient systems for detecting and identifying such elements. While extensive literature and works exist on object detection based on RGB data, it is also critical to recognize the limits of such modality when applied to UAVs detection. Detecting drones indeed poses several challenges such as fast-moving objects and scenes with a high dynamic range or, even worse, scarce illumination levels. Neuromorphic cameras, on the other hand, can retain precise and rich spatio-temporal information in situations that are challenging for RGB cameras. They are resilient to both high-speed moving objects and scarce illumination settings, while prone to suffer a rapid loss of information when the objects in the scene are static. In this context, we present a novel model for integrating both domains together, leveraging multimodal data to take advantage of the best of both worlds. To this end, we also release NeRDD (Neuromorphic-RGB Drone Detection), a novel spatio-temporally synchronized Event-RGB Drone detection dataset of more than 3.5 hours of multimodal annotated recordings., Comment: Accepted at NeVi Workshop at ECCV24

We give a rigorous derivation of the Hartree equation for the many-body dynamics of pseudo-relativistic Fermi systems at high density $\varrho \gg 1$, on arbitrarily large domains, at zero temperature. With respect to previous works, we show that the many-body evolution can be approximated by the Hartree dynamics locally, proving convergence of the expectation of observables that are supported in regions with fixed volume, independent of $\varrho$. The result applies to initial data describing fermionic systems at equilibrium confined in arbitrarily large domains, under the assumption that a suitable local Weyl-type estimate holds true. The proof relies on the approximation of the initial data through positive temperature quasi-free states, that satisfy strong local semiclassical bounds, which play a key role in controlling the growth of the local excitations of the quasi-free state along the many-body dynamics., Comment: 58 pages, no figures

Prototypical part learning is emerging as a promising approach for making semantic segmentation interpretable. The model selects real patches seen during training as prototypes and constructs the dense prediction map based on the similarity between parts of the test image and the prototypes. This improves interpretability since the user can inspect the link between the predicted output and the patterns learned by the model in terms of prototypical information. In this paper, we propose a method for interpretable semantic segmentation that leverages multi-scale image representation for prototypical part learning. First, we introduce a prototype layer that explicitly learns diverse prototypical parts at several scales, leading to multi-scale representations in the prototype activation output. Then, we propose a sparse grouping mechanism that produces multi-scale sparse groups of these scale-specific prototypical parts. This provides a deeper understanding of the interactions between multi-scale object representations while enhancing the interpretability of the segmentation model. The experiments conducted on Pascal VOC, Cityscapes, and ADE20K demonstrate that the proposed method increases model sparsity, improves interpretability over existing prototype-based methods, and narrows the performance gap with the non-interpretable counterpart models. Code is available at github.com/eceo-epfl/ScaleProtoSeg., Comment: 8 pages, 5 figures, 4 tables

Software-defined networking (SDN) is a centralized, dynamic, and programmable network management technology that enables flexible traffic control and scalability. SDN facilitates network administration through a centralized view of the underlying physical topology; tampering with this topology view can result in catastrophic damage to network management and security. To underscore this issue, we introduce Marionette, a new topology poisoning technique that manipulates OpenFlow link discovery packet forwarding to alter topology information. Our approach exposes an overlooked yet widespread attack vector, distinguishing itself from traditional link fabrication attacks that tamper, spoof, or relay discovery packets at the data plane. Unlike localized attacks observed in existing methods, our technique introduces a globalized topology poisoning attack that leverages control privileges. Marionette implements a reinforcement learning algorithm to compute a poisoned topology target, and injects flow entries to achieve a long-lived stealthy attack. Our evaluation shows that Marionette successfully attacks five open-source controllers and nine OpenFlow-based discovery protocols. Marionette overcomes the state-of-the-art topology poisoning defenses, showcasing a new class of topology poisoning that initiates on the control plane. This security vulnerability was ethically disclosed to OpenDaylight, and CVE-2024-37018 has been assigned., Comment: 17 pages, 12 figures, CCS 2024

We provide an approach enabling one to employ physics-informed neural networks (PINNs) for uncertainty quantification. Our approach is applicable to systems where observations are scarce (or even lacking), these being typical situations associated with subsurface water bodies. Our novel physics-informed neural network under uncertainty (PINN-UU) integrates the space-time domain across which processes take place and uncertain parameter spaces within a unique computational domain. PINN-UU is then trained to satisfy the relevant physical principles (e.g., mass conservation) in the defined input domain. We employ a stage training approach via transfer learning to accommodate high-dimensional solution spaces. We demonstrate the effectiveness of PINN-UU in a scenario associated with reactive transport in porous media, showcasing its reliability, efficiency, and applicability to sensitivity analysis. PINN-UU emerges as a promising tool for robust uncertainty quantification, with broad applicability to groundwater systems. As such, it can be considered as a valuable alternative to traditional methods such as multi-realization Monte Carlo simulations based on direct solvers or black-box surrogate models., Comment: 25 pages, 9 figures

For plane frameworks with reflection or rotational symmetries, where the group action is not necessarily free on the vertex set, we introduce a phase-symmetric orbit rigidity matrix for each irreducible representation of the group. We then use these generalised orbit rigidity matrices to provide necessary conditions for infinitesimal rigidity for frameworks that are symmetric with a cyclic group that acts freely or non-freely on the vertices. Moreover, for the reflection, the half-turn, and the three-fold rotational group in the plane, we establish complete combinatorial characterisations of symmetry-generic infinitesimally rigid frameworks. This extends well-known characterisations for these groups to the case when the group action is not necessarily free on the vertices. The presence of vertices that are fixed by non-trivial group elements requires the introduction of generalised versions of group-labelled quotient graphs leads to more refined types of combinatorial sparsity counts for characterising symmetry-generic infinitesimal rigidity., Comment: 36 pages excluding references, 17 figures

We show experimentally that an effect of motion of ions, observed in a plasma-based accelerator, depends inversely on the plasma ion mass. The effect appears within a single wakefield event and manifests itself as a bunch tail, occurring only when sufficient motion of ions suppresses wakefields. Wakefields are driven resonantly by multiple bunches, and simulation results indicate that the ponderomotive force causes the motion of ions. In this case, the effect is also expected to depend on the amplitude of the wakefields, experimentally confirmed through variations in the drive bunch charge.

Plasma wakefields offer high acceleration gradients, orders of magnitude larger than conventional RF accelerators. However, the achievable luminosity remains relatively low, typically limited by repetition rate and the charge accelerated per shot. In this work, we show that a train of drive bunches can be harnessed to accelerate multiple witness bunches in a single shot. We demonstrate that periodically loading the wakefields removes the limit on the energy transfer from the drive beam to the plasma, which allows the luminosity to be increased. Proof-of-concept simulations for the AWAKE scheme are carried out to demonstrate the technique, achieving a doubling of the accelerated charge while exploiting only a fraction of the drive train., Comment: 7 pages, 4 figures