arXiv subjects

Brian Moser

Publications and source records attributed to Brian Moser.

12 recordsLinked to original sources

arXiv2026

Hyperspherical Forward-Forward with Prototypical Representations

The Forward-Forward (FF) algorithm presents a compelling, bio-inspired alternative to backpropagation. However, while efficient in training, it has a computationally prohibitive inference process that requires a separate forward pass for every class that is evaluated. In this work, we introduce the Hyperspherical Forward-Forward (HFF), a novel reformulation that resolves this critical bottleneck. Our core innovation is to reframe the local objective of each layer from a binary goodness-of-fit task to a direct multi-class classification problem within a hyperspherical feature space. We achieve this by learning a set of class-specific, unit-norm prototypes that act as geometric anchors and implicit negatives. This architectural innovation preserves the benefits of local training while enabling weight update and inference in a single forward pass, making it >40x faster than the original FF algorithm. Our method is simple to implement, scales effectively to modern convolutional architectures, and achieves superior accuracy on standard image classification benchmarks, closing the gap with backpropagation. Most notably, we are among the first greedy local-learning methods to report over 25% top-1 accuracy on ImageNet-1k, and 65.96% with transfer learning.

cs.LG↗

arXiv2026

Constraining the Higgs potential using multi-Higgs production

The Higgs self-couplings remain only weakly constrained by current Large Hadron Collider (LHC) measurements, leaving ample room for physics beyond the Standard Model that could modify the structure of the Higgs potential. Multi-Higgs production processes provide a particularly sensitive probe of deviations in both the Higgs trilinear and quartic self-couplings. In this note, we summarize the current status of next-to-leading-order electroweak (EW) corrections to double-Higgs production computed within the Standard Model Effective Field Theory and Higgs Effective Field Theory frameworks, emphasizing how these calculations introduce sensitivity to the Higgs self-couplings beyond what is accessible at leading order. We discuss the key conceptual and technical differences between the two effective field theory approaches, including their treatment of higher-dimensional operators, renormalization procedures, and the structure of EW~two-loop amplitudes. Despite these differences, both approaches yield broadly consistent constraints, illustrating the complementarity of double- and triple-Higgs measurements. With the high-luminosity LHC and future high-energy colliders on the horizon, these developments and further advances provide an essential foundation for extracting increasingly precise information on the dynamics of EW symmetry breaking.

hep-ph↗

arXiv2025

HandyLabel: Towards Post-Processing to Real-Time Annotation Using Skeleton Based Hand Gesture Recognition

The success of machine learning is deeply linked to the availability of high-quality training data, yet retrieving and manually labeling new data remains a time-consuming and error-prone process. Traditional annotation tools, such as Label Studio, often require post-processing, where users label data after it has been recorded. Post-processing is highly time-consuming and labor-intensive, especially with large datasets, and may lead to erroneous annotations due to the difficulty of subjects' memory tasks when labeling cognitive activities such as emotions or comprehension levels. In this work, we introduce HandyLabel, a real-time annotation tool that leverages hand gesture recognition to map hand signs for labeling. The application enables users to customize gesture mappings through a web-based interface, allowing for real-time annotations. To ensure the performance of HandyLabel, we evaluate several hand gesture recognition models on an open-source hand sign (HaGRID) dataset, with and without skeleton-based preprocessing. We discovered that ResNet50 with preprocessed skeleton-based images performs an F1-score of 0.923. To validate the usability of HandyLabel, a user study was conducted with 46 participants. The results suggest that 88.9% of participants preferred HandyLabel over traditional annotation tools.

cs.HC↗

arXiv2025

ForAug: Mitigating Biases in Image Classification via Controlled Image Compositions

Large-scale image classification datasets exhibit strong compositional biases: objects tend to be centered, appear at characteristic scales, and co-occur with class-specific context. By exploiting such biases, models attain high in-distribution accuracy but remain fragile under distribution shifts. To address this issue, we introduce ForAug, a controlled composition augmentation scheme that factorizes each training image into a foreground object and a background and recombines them to explicitly manipulate object position, object scale, and background identity. ForAug uses off-the-shelf segmentation and inpainting models to (i) extract the foreground and synthesize a neutral background, and (ii) paste the foreground onto diverse neutral backgrounds before applying standard strong augmentation policies. Compared to conventional augmentations and content-mixing methods, our factorization provides direct control knobs that break foreground-background correlations. Across 10 architectures, ForAug improves ImageNet top-1 accuracy by up to 6 percentage points (p.p.) and yields gains of up to 7.3 p.p. on fine-grained downstream datasets. Moreover, the same control knobs enable targeted diagnostic tests: we quantify background reliance, foreground focus, center bias, and size bias via controlled background swaps and position/scale sweeps, and show that training with ForAug substantially reduces these shortcut behaviors and significantly increases accuracy on standard distribution-shift benchmarks by up to $19$ p.p. Our code and dataset are publicly available at https://github.com/tobna/ForAug.

cs.CV↗

arXiv2024

Measuring the ATLAS ITk Pixel Detector Material via Multiple Scattering of Positrons at the CERN PS

The ITk is a new silicon tracker for the ATLAS experiment designed to increase detector resolution, readout capacity, and radiation hardness, in preparation for the larger number of simultaneous proton-proton interactions at the High Luminosity LHC. This paper presents the first direct measurement of the material budget of an ATLAS ITk pixel module, performed at a testbeam at the CERN Proton Synchrotron via the multiple scattering of low energy positrons within the module volume. Using a four plane telescope of thin monolithic pixel detectors from the MALTA collaboration, scattering datasets were recorded at a beam energy of $1.2\,\text{GeV}$. Kink angle distributions were extracted from tracks derived with and without information from the ITk pixel module, and were fit to extract the RMS scattering angle, which was converted to a fractional radiation length $x/X_0$. The average $x/X_0$ across the module was measured as $[0.89 \pm 0.01 \text{ (resolution)} \pm 0.01 \text{ (subtraction)} \pm 0.08 \text{ (beam momentum band)}]\%$, which agrees within uncertainties with an estimate of $0.88\%$ derived from material component expectations.

physics.ins-det↗

arXiv2024

FedAD-Bench: A Unified Benchmark for Federated Unsupervised Anomaly Detection in Tabular Data

The emergence of federated learning (FL) presents a promising approach to leverage decentralized data while preserving privacy. Furthermore, the combination of FL and anomaly detection is particularly compelling because it allows for detecting rare and critical anomalies (usually also rare in locally gathered data) in sensitive data from multiple sources, such as cybersecurity and healthcare. However, benchmarking the performance of anomaly detection methods in FL environments remains an underexplored area. This paper introduces FedAD-Bench, a unified benchmark for evaluating unsupervised anomaly detection algorithms within the context of FL. We systematically analyze and compare the performance of recent deep learning anomaly detection models under federated settings, which were typically assessed solely in centralized settings. FedAD-Bench encompasses diverse datasets and metrics to provide a holistic evaluation. Through extensive experiments, we identify key challenges such as model aggregation inefficiencies and metric unreliability. We present insights into FL's regularization effects, revealing scenarios in which it outperforms centralized approaches due to its inherent ability to mitigate overfitting. Our work aims to establish a standardized benchmark to guide future research and development in federated anomaly detection, promoting reproducibility and fair comparison across studies.

cs.LG↗

arXiv2024

HHH Whitepaper

We here report on the progress of the HHH Workshop, that took place in Dubrovnik in July 2023. After the discovery of a particle that complies with the properties of the Higgs boson of the Standard Model, all Standard Model (SM) parameters are in principle determined. However, in order to verify or falsify the model, the full form of the potential has to be determined. This includes the measurement of the triple and quartic scalar couplings. We here report on ongoing progress of measurements for multi-scalar final states, with an emphasis on three SM-like scalar bosons at 125 GeV, but also mentioning other options. We discuss both experimental progress and challenges as well as theoretical studies and models that can enhance such rates with respect to the SM predictions

hep-ph↗

arXiv2024

Addendum to: Constraints on the quartic Higgs self-coupling from double-Higgs production at future hadron colliders

We study inclusive double-Higgs boson production at the LHC and at the HL-LHC including variations of the trilinear and of the quartic Higgs boson self-couplings at next-to-leading order (NLO) in QCD with full top quark mass dependence. Our results include the two-loop contributions to the $gg \rightarrow HH$ amplitudes that involve a modified $h_4$ vertex calculated in arXiv:1810.04665. We present results at 13, 13.6 and 14 TeV centre-of-mass energies. The implementation of the calculation is made publicly available in the POWHEG-BOX-V2 Monte Carlo framework.

hep-ph↗

arXiv2023

DartsReNet: Exploring new RNN cells in ReNet architectures

We present new Recurrent Neural Network (RNN) cells for image classification using a Neural Architecture Search (NAS) approach called DARTS. We are interested in the ReNet architecture, which is a RNN based approach presented as an alternative for convolutional and pooling steps. ReNet can be defined using any standard RNN cells, such as LSTM and GRU. One limitation is that standard RNN cells were designed for one dimensional sequential data and not for two dimensions like it is the case for image classification. We overcome this limitation by using DARTS to find new cell designs. We compare our results with ReNet that uses GRU and LSTM cells. Our found cells outperform the standard RNN cells on CIFAR-10 and SVHN. The improvements on SVHN indicate generalizability, as we derived the RNN cell designs from CIFAR-10 without performing a new cell search for SVHN.

cs.CV↗

arXiv2023

Waving Goodbye to Low-Res: A Diffusion-Wavelet Approach for Image Super-Resolution

This paper presents a novel Diffusion-Wavelet (DiWa) approach for Single-Image Super-Resolution (SISR). It leverages the strengths of Denoising Diffusion Probabilistic Models (DDPMs) and Discrete Wavelet Transformation (DWT). By enabling DDPMs to operate in the DWT domain, our DDPM models effectively hallucinate high-frequency information for super-resolved images on the wavelet spectrum, resulting in high-quality and detailed reconstructions in image space. Quantitatively, we outperform state-of-the-art diffusion-based SISR methods, namely SR3 and SRDiff, regarding PSNR, SSIM, and LPIPS on both face (8x scaling) and general (4x scaling) SR benchmarks. Meanwhile, using DWT enabled us to use fewer parameters than the compared models: 92M parameters instead of 550M compared to SR3 and 9.3M instead of 12M compared to SRDiff. Additionally, our method outperforms other state-of-the-art generative methods on classical general SR datasets while saving inference time. Finally, our work highlights its potential for various applications.

cs.CV↗

arXiv2022

Hitchhiker's Guide to Super-Resolution: Introduction and Recent Advances

With the advent of Deep Learning (DL), Super-Resolution (SR) has also become a thriving research area. However, despite promising results, the field still faces challenges that require further research e.g., allowing flexible upsampling, more effective loss functions, and better evaluation metrics. We review the domain of SR in light of recent advances, and examine state-of-the-art models such as diffusion (DDPM) and transformer-based SR models. We present a critical discussion on contemporary strategies used in SR, and identify promising yet unexplored research directions. We complement previous surveys by incorporating the latest developments in the field such as uncertainty-driven losses, wavelet networks, neural architecture search, novel normalization methods, and the latests evaluation techniques. We also include several visualizations for the models and methods throughout each chapter in order to facilitate a global understanding of the trends in the field. This review is ultimately aimed at helping researchers to push the boundaries of DL applied to SR.

cs.CV↗

arXiv2022

Less is More: Proxy Datasets in NAS approaches

Neural Architecture Search (NAS) defines the design of Neural Networks as a search problem. Unfortunately, NAS is computationally intensive because of various possibilities depending on the number of elements in the design and the possible connections between them. In this work, we extensively analyze the role of the dataset size based on several sampling approaches for reducing the dataset size (unsupervised and supervised cases) as an agnostic approach to reduce search time. We compared these techniques with four common NAS approaches in NAS-Bench-201 in roughly 1,400 experiments on CIFAR-100. One of our surprising findings is that in most cases we can reduce the amount of training data to 25\%, consequently reducing search time to 25\%, while at the same time maintaining the same accuracy as if training on the full dataset. Additionally, some designs derived from subsets out-perform designs derived from the full dataset by up to 22 p.p. accuracy.

cs.LG↗