Search

Recent approaches in machine learning often solve a task using a composition of multiple models or agentic architectures. When targeting a composed system with adversarial attacks, it might not be computationally or informationally feasible to train an end-to-end proxy model or a proxy model for every component of the system. We introduce a method to craft an adversarial attack against the overall multi-model system when we only have a proxy model for the final black-box model, and when the transformation applied by the initial models can make the adversarial perturbations ineffective. Current methods handle this by applying many copies of the first model/transformation to an input and then re-use a standard adversarial attack by averaging gradients, or learning a proxy model for both stages. To our knowledge, this is the first attack specifically designed for this threat model and our method has a substantially higher attack success rate (80% vs 25%) and contains 9.4% smaller perturbations (MSE) compared to prior state-of-the-art methods. Our experiments focus on a supervised image pipeline, but we are confident the attack will generalize to other multi-model settings [e.g. a mix of open/closed source foundation models], or agentic systems, Comment: 11 pages, 2 figures

The classification of jets as quark- versus gluon-initiated is an important yet challenging task in the analysis of data from high-energy particle collisions and in the search for physics beyond the Standard Model. The recent integration of deep neural networks operating on low-level detector information has resulted in significant improvements in the classification power of quark/gluon jet tagging models. However, the improved power of such models trained on simulated samples has come at the cost of reduced interpretability, raising concerns about their reliability. We elucidate the physics behind quark/gluon jet classification decisions by comparing the performance of networks with and without constraints of infrared and collinear safety, and identify the nature of the unsafe information by revealing the energy and angular dependence of the learned models. This in turn allows us to approximate the performance of the low-level networks (by 99\% or higher) using equivalent sets of interpretable high-level observables, which can be used to probe the fidelity of the simulated samples and define systematic uncertainties.

Distinguishing between prompt muons produced in heavy boson decay and muons produced in association with heavy-flavor jet production is an important task in analysis of collider physics data. We explore whether there is information available in calorimeter deposits that is not captured by the standard approach of isolation cones. We find that convolutional networks and particle-flow networks accessing the calorimeter cells surpass the performance of isolation cones, suggesting that the radial energy distribution and the angular structure of the calorimeter deposits surrounding the muon contain unused discrimination power. We assemble a small set of high-level observables which summarize the calorimeter information and close the performance gap with networks which analyze the calorimeter cells directly. These observables are theoretically well-defined and can be studied with collider data., Comment: Version accepted for publication at JHEP

QCD-jets at the LHC are described by simple physics principles. We show how super-resolution generative networks can learn the underlying structures and use them to improve the resolution of jet images. We test this approach on massless QCD-jets and on fat top-jets and find that the network reproduces their main features even without training on pure samples. In addition, we show how a slim network architecture can be constructed once we have control of the full network performance., Comment: 25 pages, 11 figures; implemented SciPost reviewer comments, clarified definitions and expanded discussion in high-level observable benchmarking subsection (section 3.3 and Fig. 7)

In the framework of three-active-neutrino mixing, the charge parity phase, the neutrino mass ordering, and the octant of $\theta_{23}$ remain unknown. The Deep Underground Neutrino Experiment (DUNE) is a next-generation long-baseline neutrino oscillation experiment, which aims to address these questions by measuring the oscillation patterns of $\nu_\mu/\nu_e$ and $\bar\nu_\mu/\bar\nu_e$ over a range of energies spanning the first and second oscillation maxima. DUNE far detector modules are based on liquid argon TPC (LArTPC) technology. A LArTPC offers excellent spatial resolution, high neutrino detection efficiency, and superb background rejection, while reconstruction in LArTPC is challenging. Deep learning methods, in particular, Convolutional Neural Networks (CNNs), have demonstrated success in classification problems such as particle identification in DUNE and other neutrino experiments. However, reconstruction of neutrino energy and final state particle momenta with deep learning methods is yet to be developed for a full AI-based reconstruction chain. To precisely reconstruct these kinematic characteristics of detected interactions at DUNE, we have developed and will present two CNN-based methods, 2-D and 3-D, for the reconstruction of final state particle direction and energy, as well as neutrino energy. Combining particle masses with the kinetic energy and the direction reconstructed by our work, the four-momentum of final state particles can be obtained. Our models show considerable improvements compared to the traditional methods for both scenarios., Comment: NuerIPS workshop proceeding

We investigate whether state-of-the-art classification features commonly used to distinguish electrons from jet backgrounds in collider experiments are overlooking valuable information. A deep convolutional neural network analysis of electromagnetic and hadronic calorimeter deposits is compared to the performance of typical features, revealing a $\approx 5\%$ gap which indicates that these lower-level data do contain untapped classification power. To reveal the nature of this unused information, we use a recently developed technique to map the deep network into a space of physically interpretable observables. We identify two simple calorimeter observables which are not typically used for electron identification, but which mimic the decisions of the convolutional network and nearly close the performance gap., Comment: 10 pages, lots of figures, v2 for submission

Generation of simulated data is essential for data analysis in particle physics, but current Monte Carlo methods are very computationally expensive. Deep-learning-based generative models have successfully generated simulated data at lower cost, but struggle when the data are very sparse. We introduce a novel deep sparse autoregressive model (SARM) that explicitly learns the sparseness of the data with a tractable likelihood, making it more stable and interpretable when compared to Generative Adversarial Networks (GANs) and other methods. In two case studies, we compare SARM to a GAN model and a non-sparse autoregressive model. As a quantitative measure of performance, we compute the Wasserstein distance ($W_p$) between the distributions of physical quantities calculated on the generated images and on the training images. In the first study, featuring images of jets in which 90% of the pixels are zero-valued, SARM produces images with $W_p$ scores that are 24-52% better than the scores obtained with other state-of-the-art generative models. In the second study, on calorimeter images in the vicinity of muons where 98% of the pixels are zero-valued, SARM produces images with $W_p$ scores that are 66-68% better. Similar observations made with other metrics confirm the usefulness of SARM for sparse data in particle physics. Original data and software will be made available upon acceptance of the manuscript from the UCI Machine Learning in Physics web portal at: http://mlphysics.ics.uci.edu/., Comment: changed templates

Sherpa is a hyperparameter optimization library for machine learning models. It is specifically designed for problems with computationally expensive, iterative function evaluations, such as the hyperparameter tuning of deep neural networks. With Sherpa, scientists can quickly optimize hyperparameters using a variety of powerful and interchangeable algorithms. Sherpa can be run on either a single machine or in parallel on a cluster. Finally, an interactive dashboard enables users to view the progress of models as they are trained, cancel trials, and explore which hyperparameter combinations are working best. Sherpa empowers machine learning practitioners by automating the more tedious aspects of model tuning. Its source code and documentation are available at https://github.com/sherpa-ai/sherpa.

Classification of jets as originating from light-flavor or heavy-flavor quarks is an important task for inferring the nature of particles produced in high-energy collisions. The large and variable dimensionality of the data provided by the tracking detectors makes this task difficult. The current state-of-the-art tools require expert data-reduction to convert the data into a fixed low-dimensional form that can be effectively managed by shallow classifiers. We study the application of deep networks to this task, attempting classification at several levels of data, starting from a raw list of tracks. We find that the highest-level lowest-dimensionality expert information sacrifices information needed for classification, that the performance of current state-of-the-art taggers can be matched or slightly exceeded by deep-network-based taggers using only track and vertex information, that classification using only lowest-level highest-dimensionality tracking information remains a difficult task for deep networks, and that adding lower-level track and vertex information to the classifiers provides a significant boost in performance compared to the state-of-the-art., Comment: 12 pages, submitted to PRD