Search

Recently, nonlinear effects have been utilized to cool a special beam represented by an annular distribution in a 2D phase space. This outcome was accomplished using an AC dipole combined with amplitude detuning generated by static nonlinear magnets. In this paper, we investigate a more realistic scenario in which a beam distribution in a 4D phase space includes the presence of a beam halo and demonstrate how the latter can be removed using nonlinear effects. The proposed approach employs high-order nonlinear AC magnets to trap into nonlinear resonances and to perform an adiabatic transport of the beam halo in phase space. Theoretical models will be formulated and examined through numerical simulations to evaluate their efficacy.

In circular particle accelerators, storage rings, or colliders, mitigating beam losses is critical to ensuring optimal performance, particularly for rings that include superconducting magnets. A thorough understanding of beam-halo dynamics is essential for this purpose. This paper presents recent results for the measurement of the nonlinear diffusion process of the beam halo at the CERN Large Hadron Collider (LHC). The novel approach used in this paper is based on the analytical framework of the Nekhoroshev theorem, which provides a functional form for the diffusion coefficient. By monitoring the beam loss signal during controlled movements of the collimator jaws, we determine the beam losses at equilibrium for various amplitudes and analyze the beam-halo distribution. Post-processing of these measurements provides the nonlinear diffusion coefficient, which is found to be in excellent agreement with the theoretical assumptions. Measurements from an experiment investigating the effectiveness of beam-beam compensation using beam-beam compensation wires also provide a direct assessment of the compensation's effectiveness on beam-tail diffusion.

Large-scale vision-language pre-trained (VLP) models (e.g., CLIP) are renowned for their versatility, as they can be applied to diverse applications in a zero-shot setup. However, when these models are used in specific domains, their performance often falls short due to domain gaps or the under-representation of these domains in the training data. While fine-tuning VLP models on custom datasets with human-annotated labels can address this issue, annotating even a small-scale dataset (e.g., 100k samples) can be an expensive endeavor, often requiring expert annotators if the task is complex. To address these challenges, we propose LatteCLIP, an unsupervised method for fine-tuning CLIP models on classification with known class names in custom domains, without relying on human annotations. Our method leverages Large Multimodal Models (LMMs) to generate expressive textual descriptions for both individual images and groups of images. These provide additional contextual information to guide the fine-tuning process in the custom domains. Since LMM-generated descriptions are prone to hallucination or missing details, we introduce a novel strategy to distill only the useful information and stabilize the training. Specifically, we learn rich per-class prototype representations from noisy generated texts and dual pseudo-labels. Our experiments on 10 domain-specific datasets show that LatteCLIP outperforms pre-trained zero-shot methods by an average improvement of +4.74 points in top-1 accuracy and other state-of-the-art unsupervised methods by +3.45 points.

The congestion formation on a urban road network is one of the key issue for the development of a sustainable mobility in the future smart cities. In this work we propose a reductionist approach studying the stationary states of a simple transport model using of a random process on a graph, where each node represents a location and the weight links give the transition rates to move from one node to another that represent the mobility demand. Each node has a finite transport capacity and a maximum load capacity and we assume that the average. In the approximation of the single step process we are able to analytically characterize the traffic load distribution on the single nodes, using a local Maximum Entropy Principle. Our results explain how the congested nodes emerge when the total traffic load increases in analogous way to a percolation transition where the appearance of a congested node is a independent random event, However, using numerical simulations, we show that in the more realistic case of the synchronous dynamics for the nodes, there are entropic forces that introduce correlation among the node state and favor the clustering of the empty and congested nodes. Our aim is to highlight universal properties of the congestion formation and, in particular, to understand the role traffic load fluctuations as a possible precursor of congestion in a transport network., Comment: 24 pages, 8 figures, regular article

In this paper, we analyze the adiabatic crossing of a resonance for Hamiltonian systems when a double-resonance condition is satisfied by the linear frequency at an elliptic fixed point. We discuss in detail the phase-space structure on a class of Hamiltonians and area-preserving maps with an elliptic fixed point in the presence of a time-dependent exciter. Various regimes have been identified and carefully studied. This study extends results obtained recently for the trapping and transport phenomena for periodically perturbed Hamiltonian systems, and it could have relevant applications in the adiabatic beam splitting in accelerator physics.

Novel-view synthesis through diffusion models has demonstrated remarkable potential for generating diverse and high-quality images. Yet, the independent process of image generation in these prevailing methods leads to challenges in maintaining multiple-view consistency. To address this, we introduce ViewFusion, a novel, training-free algorithm that can be seamlessly integrated into existing pre-trained diffusion models. Our approach adopts an auto-regressive method that implicitly leverages previously generated views as context for the next view generation, ensuring robust multi-view consistency during the novel-view generation process. Through a diffusion process that fuses known-view information via interpolated denoising, our framework successfully extends single-view conditioned models to work in multiple-view conditional settings without any additional fine-tuning. Extensive experimental results demonstrate the effectiveness of ViewFusion in generating consistent and detailed novel views., Comment: CVPR2024,homepage:https://wi-sc.github.io/ViewFusion.github.io/

Diffusion on complex networks is a convenient framework to describe a great variety of transport systems. Failure phenomena in a link of the network may simulate the presence of a break or a congestion effect in the system. A real time detection of failures can mitigate their effect and allow to optimize the control procedures on the transport network. The main objective of this work is to provide a dimensionality reduction technique for a transport network where a diffusive dynamics takes place, to detect presence of a failure by a limited number of observations. Our approach is based on the susceptibility response of the network state under random perturbations of the link weights. The correlations among the nodes fluctuations is exploited in order to provide the clustering procedure. The network dimensionality is therefore reduced introducing `representative nodes' for each cluster and generating a reduced network model, whose dynamical state is detected by the observations. We realize a failure identification procedure for the whole network, studying the dynamics of the coarse-grained network. The localization efficiency of the proposed clustering algorithm, averaging over all possible single-edge failures, is compared with traditional structure-based clustering using different graph configurations. We show that the proposed clustering algorithm is more sensitive than traditional clustering techniques to detect link failure with higher stationary fluxes., Comment: 31 pages

Time-bin encoding has been widely used for implementing quantum key distribution (QKD) on optical fiber channels due to its robustness with respect to drifts introduced by the optical fiber. However, due to the use of interferometric structures, achieving stable and low intrinsic Quantum Bit Error rate (QBER) in time-bin systems can be challenging. A key device for decoy-state prepare & measure QKD is represented by the state encoder, that must generate low-error and stable states with different values of mean photon number. Here we propose the MacZac (Mach-Zehder-Sagnac), a time-bin encoder with ultra-low intrinsic QBER (<2e-5) and high stability. The device is based on nested Sagnac and Mach-Zehnder interferometers and uses a single phase modulator for both decoy and state preparation, greatly simplifying the optical setup. The encoder does not require any active compensation or feedback system and it can be scaled for the generation of states with arbitrary dimension. We experimentally realized and tested the device performances as a stand alone component and in a complete QKD experiments. Thanks to the capacity to combine extremely low QBER, high stability and experimental simplicity the proposed device can be used as a key building block for future high-performance, low-cost QKD systems., Comment: 12 pages, 14 figures

This work aims at investigating the optical transmission system needed for such lightweight sail, taking into account the physical constraints of such unprecedented link and focusing on the optimal scheme for the optical signal emission. In particular, the optical signal is distributed to several emitters on the sail. The light diffraction resulting from the pattern of the emitters acting coherently determines the characteristics of the whole beam transmitted by the sail and of the received signal on the Earth. The performance of the digital communication system using pulse position modulation (PPM) can be assessed and channel coding schemes are proposed. We are using the paradigm for which the entire sail communication system is described as a Tree-of-light: the detectors, CPU, memory and laser transmitter are the central unit, representing the trunk of the tree. The branches of the tree are the waveguides, directed to the sail surface. By means of multimode splitters, the signal is further distributed via the petioles to the emitters, the leaves, realized by grating couplers (GCs), on which this work is more focused.

Diffusion models (DMs) can generate realistic images with text guidance using large-scale datasets. However, they demonstrate limited controllability in the output space of the generated images. We propose a novel learning method for text-guided image editing, namely \texttt{iEdit}, that generates images conditioned on a source image and a textual edit prompt. As a fully-annotated dataset with target images does not exist, previous approaches perform subject-specific fine-tuning at test time or adopt contrastive learning without a target image, leading to issues on preserving the fidelity of the source image. We propose to automatically construct a dataset derived from LAION-5B, containing pseudo-target images with their descriptive edit prompts given input image-caption pairs. This dataset gives us the flexibility of introducing a weakly-supervised loss function to generate the pseudo-target image from the latent noise of the source image conditioned on the edit prompt. To encourage localised editing and preserve or modify spatial structures in the image, we propose a loss function that uses segmentation masks to guide the editing during training and optionally at inference. Our model is trained on the constructed dataset with 200K samples and constrained GPU resources. It shows favourable results against its counterparts in terms of image fidelity, CLIP alignment score and qualitatively for editing both generated and real images.

The efficient detection of chaotic behavior in orbits of a complex dynamical system is an active domain of research. Several indicators have been proposed in the past, and new ones have recently been developed in view of improving the performance of chaos detection by means of numerical simulations. The challenge is to predict chaotic behavior based on the analysis of orbits of limited length. In this paper, the performance analysis of past and recent indicators of chaos, in terms of predictive power, is carried out in detail using the dynamical system characterized by a symplectic H\'enon-like cubic polynomial map.

In a recent paper, the adiabatic theory of Hamiltonian systems was successfully applied to study the crossing of the linear coupling resonance, $Q_x-Q_y=0$. A detailed explanation of the well-known phenomena that occur during the resonance-crossing process, such as emittance exchange and its dependence on the adiabaticity of the process was obtained. In this paper, we consider the crossing of the resonance of nonlinear coupling $2 Q_x -2 Q_y = 0$ using the same theoretical framework. We perform the analysis using a Hamiltonian model in which the nonlinear coupling resonance is excited and the corresponding dynamics is studied in detail, in particular looking at the phase-space topology and its evolution, in view of characterizing the emittance exchange phenomena. The theoretical results are then tested using a symplectic map. Thanks to this approach, scaling laws of general interest for applications are derived.

In this paper, new results concerning the phenomenon of adiabatic trapping into resonance for a class of quasi-integrable maps with a time-dependent exciter are presented and discussed in detail. The applicability of the results about trapping efficiency for Hamiltonian systems to the maps considered is proven by using perturbation theory. This allows determining explicit scaling laws for the trapping properties. These findings represent a generalization of previous results obtained for the case of quasi-integrable maps with parametric modulation as well as an extension of the work by Neishtadt \textit{et al.} on a restricted class of quasi-integrable systems with time-dependent exciters.

In recent years, intense efforts have been devoted to studying how nonlinear effects can be used to shape the transverse beam distribution by means of an adiabatic crossing of nonlinear resonances. By this approach, it is possible to split the beams in the transverse plane, so that the initial single-Gaussian beam is divided into several distinct distributions. This is at the heart of the multiturn extraction process that is successfully in operation at the CERN Proton Synchrotron. Nonlinear effects can also be used to cool a beam by acting on its transverse beam distribution. In this paper, we present and discuss the special case of a beam with an annular distribution, showing how its emittance can be effectively reduced by means of properly devised manipulations based on nonlinear effects.

Long-form video understanding requires designing approaches that are able to temporally localize activities or language. End-to-end training for such tasks is limited by the compute device memory constraints and lack of temporal annotations at large-scale. These limitations can be addressed by pre-training on large datasets of temporally trimmed videos supervised by class annotations. Once the video encoder is pre-trained, it is common practice to freeze it during fine-tuning. Therefore, the video encoder does not learn temporal boundaries and unseen classes, causing a domain gap with respect to the downstream tasks. Moreover, using temporally trimmed videos prevents to capture the relations between different action categories and the background context in a video clip which results in limited generalization capacity. To address these limitations, we propose a novel post-pre-training approach without freezing the video encoder which leverages language. We introduce a masked contrastive learning loss to capture visio-linguistic relations between activities, background video clips and language in the form of captions. Our experiments show that the proposed approach improves the state-of-the-art on temporal action localization, few-shot temporal action localization, and video language grounding tasks., Comment: 18 pages, 4 figures

Circular particle accelerators at the energy frontier are based on superconducting magnets that are extremely sensitive to beam losses as these might induce quenches, i.e.\ transitions to the normal-conducting state. Furthermore, the energy stored in the circulating beam is so large that hardware integrity is put in serious danger, and machine protection becomes essential for reaching the nominal accelerator performance. In this challenging context, the beam halo becomes a potential source of performance limitations and its dynamics needs to be understood in detail to assess whether it could be an issue for the accelerator. In this paper, we discuss in detail a novel framework, based on a diffusive approach, to model beam-halo dynamics. The functional form of the optimal estimate of the perturbative series, as given by Nekhoroshev's theorem, is used to provide the functional form of the action diffusion coefficient. The goal is to propose an effective model for the beam-halo dynamics and to devise an efficient experimental procedure to obtain an accurate measurement of the diffusion coefficient.

Controlling non-linear effects in the transverse dynamics of charged particle beams in circular accelerators opens new possibilities for controlling some of the beam properties. Beam splitting by crossing a stable 1D non-linear resonance is part of the routine operation of the CERN Proton Synchrotron. The beam undergoes trapping and transport inside stable islands created in the horizontal plane to allow multi-turn extraction towards the Super Proton Synchrotron, where the beam is used for fixed-target experiments. This process acts only on the horizontal beam emittance, inducing a reduction of its initial value. In this paper, we present a generalisation of this approach, in which both transverse planes are affected by the proposed technique. We will discuss in detail how to manipulate the transverse emittances by means of a controlled crossing of a 2D non-linear resonance. The novel technique will be presented by discussing the theoretical analysis of a Hamiltonian model, as well as simulating the performance of the proposed manipulation using a more realistic non-linear symplectic map.

Current technological progress is driving Quantum Key Distribution towards a commercial and world widescale expansion. Its capability to deliver unconditionally secure communication will be a fundamental feature in the next generations of telecommunication networks. Nevertheless, demonstrations of QKD implementation in a real operating scenario and their coexistence with the classical telecom infrastructure are of fundamental importance for reliable exploitation. Here we present a Quantum Key Distribution application implemented overa classical fiber-based infrastructure. By exploiting just a single fiber cable for both the quantum and the classical channel and by using a simplified receiver scheme with just one single-photon detector, we demonstrate the feasibility of low-cost and ready-to-use Quantum Key Distribution systems compatible with standard classical infrastructure., Comment: 5 pages, 5 figures

The statistical properties of human mobility have been studied in the framework of complex systems physics. Taking advantage from the new datasets made available by the information and communication technologies, the distributions of mobility path lengths and of trip duration have been considered to discover the fingerprints of complexity characters, but the role of the different transportation means on the statistical properties of urban mobility has not been studied in deep. In this paper we cope with the problem of pointing out the existence of universal features for different type of individual mobility: pedestrian, cycling and vehicular urban mobility. In particular, we propose the use of travel time as universal 'energy' for the mobility and we define a simple survival model that explains the travel time distribution of the different types of mobility. the analysis is performed in the metropolitan area of Bologna (Italy), where GPS datasets were available on individual trips using different transport means. Our results could suggest how to plan the different transportation networks to realize a multimodal mobility compatibly with the citizens propensities to use the different transport means.

In circular accelerators, crossing the linear coupling resonance induces the exchange of the transverse emittances, provided the process is adiabatic. This has been considered in some previous works, where the description of the phenomenon has been laid down, and, more recently, where a possible explanation of the numerical results has been proposed. In this paper, we introduce a theoretical framework to analyze the crossing process, based on the theory of adiabatic invariance for Hamiltonian mechanics, which explains in detail various features of the emittance exchange process.

Cross-modal recipe retrieval has recently gained substantial attention due to the importance of food in people's lives, as well as the availability of vast amounts of digital cooking recipes and food images to train machine learning models. In this work, we revisit existing approaches for cross-modal recipe retrieval and propose a simplified end-to-end model based on well established and high performing encoders for text and images. We introduce a hierarchical recipe Transformer which attentively encodes individual recipe components (titles, ingredients and instructions). Further, we propose a self-supervised loss function computed on top of pairs of individual recipe components, which is able to leverage semantic relationships within recipes, and enables training using both image-recipe and recipe-only samples. We conduct a thorough analysis and ablation studies to validate our design choices. As a result, our proposed method achieves state-of-the-art performance in the cross-modal recipe retrieval task on the Recipe1M dataset. We make code and models publicly available., Comment: CVPR 2021