arXiv subjects

Srinandan Dasmahapatra

Publications and source records attributed to Srinandan Dasmahapatra.

At least 19 recordsLinked to original sources

arXiv2026

Hunting the Unseen: Deep Learning Analysis for Semi-Visible Jet Tagging

Semi-Visible Jets (SVJs) constitute a distinctive collider signature of strongly interacting dark sectors, embedding Dark Matter candidates, wherein jets contain both visible Standard Model objects and invisible dark hadrons, giving rise to correlated jet activity and missing transverse momentum. In this work, we investigate SVJs produced through a heavy Z' mediator and perform an study over a representative set of benchmark scenarios spanning different mediator masses and dark sector parameters in the context of so-called Hidden Valley Models. To characterise the signal, we combine global event kinematics with jet substructure observables, including the primary Lund Jet Plane (LJP), the two-point energy correlation, angularity, and charged hadron multiplicity. These representations are used to train five Deep Learning classifiers for SVJ vs standard jet discrimination: a Vision Transformer operating on LJP images, a JetLOV network based on a hierarchical clustering tree, a Multi-Layer Perceptron using high level observables, and two multimodal networks that combine the image-based or hierarchical representations of the radiation pattern with the high jet-level observables. This enables a direct combination of global kinematics, radiation patterns, and jet clustering structure. We find that global kinematic observables outperform the LJP and hierarchical jet representations, with the latter providing stronger discrimination than LJP images. Combining these complementary representations with global kinematics yields the best overall performance. More broadly, this study shows that unlocking the full discovery potential of SVJs would benefit from going beyond global kinematics to exploit the rich information encoded in their internal structure, providing a benchmark for future searches at the Large Hadron Collider.

hep-ph↗

arXiv2026

Understanding Imbalanced Forgetting in Rehearsal-Based Class-Incremental Learning

Neural networks suffer from catastrophic forgetting in class-incremental learning (CIL) settings. Rehearsal$\unicode{x2013}$replaying a subset of past samples$\unicode{x2013}$is a well-established mitigation strategy. However, recent results suggest that, despite balanced rehearsal allocation, some classes are forgotten substantially more than others. Despite its relevance, this imbalanced forgetting phenomenon remains underexplored. This work shows that imbalanced forgetting arises systematically and severely in rehearsal-based CIL and investigates it extensively. Specifically, we construct, from a principled analysis, three last-layer coefficients that capture different gradient-level sources of interference affecting each past class during an incremental step. We then demonstrate that, together, they reliably predict how past classes will rank in terms of forgetting at the end of that step. While predictive performance alone does not establish causality, these results support the interpretation of the coefficients as a plausible mechanistic account linking last-layer gradient-level interactions during training to class-level forgetting outcomes. Notably, one coefficient$\unicode{x2013}$capturing self-induced interference$\unicode{x2013}$emerges as the strongest predictor, with controlled experiments providing evidence consistent with this coefficient being influenced by the new-class interference coefficient. Overall, our findings provide valuable insights and suggest promising directions for mitigating imbalanced forgetting by reducing class-wise disparities in the identified sources of interference.

cs.LG↗

arXiv2026

Mini-Batch Class Composition Bias in Link Prediction

Prior work on node classification has shown that Graph Neural Networks (GNNs) can learn representations that transfer across graphs, when underlying graph properties are shared. For a fixed graph, one would then expect GNNs trained for link prediction to learn a representation consistent with that learnt for node classification. We show this intuition does not hold in the general case. Instead, we find popular link prediction models can learn a trivial mini-batch dependent heuristic, enabled by batch-normalisation layers, to solve the edge classification task. When correcting for this, we observe increased alignment of the network representation with node-class relevant features, suggesting the network has learnt a graph representation that better aligns with the underlying graph's properties. Our findings suggest that standard link prediction training may be leading us to overestimate link predictors' ability to learn a generalised representation of a graph that is consistent across tasks.

cs.LG↗

arXiv2025

Bayesian Active Search on Parameter Space: a 95 GeV Spin-0 Resonance in the ($B-L$)SSM

In the attempt to explain possible data anomalies from collider experiments in terms of New Physics (NP) models, computationally expensive scans over their parameter spaces are typically required in order to match theoretical predictions to experimental observations. Under the assumption that anomalies seen at a mass of about 95 GeV by the Large Electron-Positron (LEP) and Large Hadron Collider (LHC) experiments correspond to a NP signal, which we attempt to interpret as a spin-0 resonance in the $(B-L)$ Supersymmetric Standard Model ($(B-L)$SSM), chosen as an illustrative example, we introduce a novel Machine Learning (ML) approach based on a multi-objective active search method, called b-CASTOR, able to achieve high sample efficiency and diversity, due to the use of probabilistic surrogate models and a volume based search policy, outperforming competing algorithms, such as those based on Markov-Chain Monte Carlo (MCMC) methods.

hep-ph↗

arXiv2024

hep-aid: A Python Library for Sample Efficient Parameter Scans in Beyond the Standard Model Phenomenology

This paper presents hep-aid, a modular Python library conceived for utilising, implementing, and developing parameter scan algorithms. Originally devised for sample-efficient, multi-objective active search approaches in computationally expensive Beyond Standard Model (BSM) phenomenology, the library currently integrates three Machine Learning (ML)-based approaches: a Constraint Active Search (CAS) algorithm, a multi-objective Active Search (AS) method (called b-CASTOR), and a self-exploration method named Machine Learning Scan (MLScan). These approaches address the challenge of multi-objective optimisation in high-dimensional BSM scenarios by employing surrogate models and strategically exploring parameter spaces to identify regions that satisfy complex objectives with fewer evaluations. Additionally, a Markov-Chain Monte Carlo method using the Metropolis-Hastings algorithm (MCMC-MH) is implemented for method comparison. The library also includes a High Energy Physics (HEP) module based on SPheno as the spectrum calculator. However, the library modules and functionalities are designed to be easily extended and used also with other external software for phenomenology. This manual provides an introduction on how to use the main functionalities of hep-aid and describes the design and structure of the library. Demonstrations based on the aforementioned parameter scan methods show that hep-aid methodologies enhance the efficiency of BSM studies, offering a versatile toolset for complex, multi-objective searches for new physics in HEP contexts exploiting advanced ML-based approaches.

hep-ph↗

arXiv2024

DBAT: Dynamic Backward Attention Transformer for Material Segmentation with Cross-Resolution Patches

The objective of dense material segmentation is to identify the material categories for every image pixel. Recent studies adopt image patches to extract material features. Although the trained networks can improve the segmentation performance, their methods choose a fixed patch resolution which fails to take into account the variation in pixel area covered by each material. In this paper, we propose the Dynamic Backward Attention Transformer (DBAT) to aggregate cross-resolution features. The DBAT takes cropped image patches as input and gradually increases the patch resolution by merging adjacent patches at each transformer stage, instead of fixing the patch resolution during training. We explicitly gather the intermediate features extracted from cross-resolution patches and merge them dynamically with predicted attention masks. Experiments show that our DBAT achieves an accuracy of 86.85%, which is the best performance among state-of-the-art real-time models. Like other successful deep learning solutions with complex architectures, the DBAT also suffers from lack of interpretability. To address this problem, this paper examines the properties that the DBAT makes use of. By analysing the cross-resolution features and the attention weights, this paper interprets how the DBAT learns from image patches. We further align features to semantic labels, performing network dissection, to infer that the proposed model can extract material-related features better than other methods. We show that the DBAT model is more robust to network initialisation, and yields fewer variable predictions compared to other models. The project code is available at https://github.com/heng-yuwen/Dynamic-Backward-Attention-Transformer.

cs.CV↗

arXiv2023

JetLOV: Enhancing Jet Tree Tagging through Neural Network Learning of Optimal LundNet Variables

Machine learning has played a pivotal role in advancing physics, with deep learning notably contributing to solving complex classification problems such as jet tagging in the field of jet physics. In this experiment, we aim to harness the full potential of neural networks while acknowledging that, at times, we may lose sight of the underlying physics governing these models. Nevertheless, we demonstrate that we can achieve remarkable results obscuring physics knowledge and relying completely on the model's outcome. We introduce JetLOV, a composite comprising two models: a straightforward multilayer perceptron (MLP) and the well-established LundNet. Our study reveals that we can attain comparable jet tagging performance without relying on the pre-computed LundNet variables. Instead, we allow the network to autonomously learn an entirely new set of variables, devoid of a priori knowledge of the underlying physics. These findings hold promise, particularly in addressing the issue of model dependence, which can be mitigated through generalization and training on diverse data sets.

hep-ph↗

arXiv2023

Probabilistic Weight Fixing: Large-scale training of neural network weight uncertainties for quantization

Weight-sharing quantization has emerged as a technique to reduce energy expenditure during inference in large neural networks by constraining their weights to a limited set of values. However, existing methods for weight-sharing quantization often make assumptions about the treatment of weights based on value alone that neglect the unique role weight position plays. This paper proposes a probabilistic framework based on Bayesian neural networks (BNNs) and a variational relaxation to identify which weights can be moved to which cluster centre and to what degree based on their individual position-specific learned uncertainty distributions. We introduce a new initialisation setting and a regularisation term which allow for the training of BNNs under complex dataset-model combinations. By leveraging the flexibility of weight values captured through a probability distribution, we enhance noise resilience and downstream compressibility. Our iterative clustering procedure demonstrates superior compressibility and higher accuracy compared to state-of-the-art methods on both ResNet models and the more complex transformer-based architectures. In particular, our method outperforms the state-of-the-art quantization method top-1 accuracy by 1.6% on ImageNet using DeiT-Tiny, with its 5 million+ weights now represented by only 296 unique values.

cs.LG↗

arXiv2023

MatSpectNet: Material Segmentation Network with Domain-Aware and Physically-Constrained Hyperspectral Reconstruction

Achieving accurate material segmentation for 3-channel RGB images is challenging due to the considerable variation in a material's appearance. Hyperspectral images, which are sets of spectral measurements sampled at multiple wavelengths, theoretically offer distinct information for material identification, as variations in intensity of electromagnetic radiation reflected by a surface depend on the material composition of a scene. However, existing hyperspectral datasets are impoverished regarding the number of images and material categories for the dense material segmentation task, and collecting and annotating hyperspectral images with a spectral camera is prohibitively expensive. To address this, we propose a new model, the MatSpectNet to segment materials with recovered hyperspectral images from RGB images. The network leverages the principles of colour perception in modern cameras to constrain the reconstructed hyperspectral images and employs the domain adaptation method to generalise the hyperspectral reconstruction capability from a spectral recovery dataset to material segmentation datasets. The reconstructed hyperspectral images are further filtered using learned response curves and enhanced with human perception. The performance of MatSpectNet is evaluated on the LMD dataset as well as the OpenSurfaces dataset. Our experiments demonstrate that MatSpectNet attains a 1.60% increase in average pixel accuracy and a 3.42% improvement in mean class accuracy compared with the most recent publication. The project code is attached to the supplementary material and will be published on GitHub.

cs.CV↗

arXiv2023

Fat $b$-Jet Analyses Using Old and New Clustering Algorithms in New Higgs Boson Searches at the LHC

We compare different jet-clustering algorithms in establishing fully hadronic final states stemming from the chain decay of a heavy Higgs state into a pair of the 125 GeV Higgs boson that decays into bottom-antibottom quark pairs. Such 4$b$ events typically give rise to boosted topologies, wherein bottom-antibottom quark pairs emerging from each 125 GeV Higgs boson tend to merge into a single, fat $b$-jet. Assuming Large Hadron Collider (LHC) settings, we illustrate how both the efficiency of selecting the multi-jet final state and the ability to reconstruct from it the masses of all Higgs bosons depend on the choice of jet-clustering algorithm and its parameter settings. We indicate the optimal choice of clustering method for the purpose of establishing such a ubiquitous Beyond the SM (BSM) signal, illustrated via a Type-II 2-Higgs Doublet Model (2HDM).

hep-ph↗

arXiv2023

ADS_UNet: A Nested UNet for Histopathology Image Segmentation

The UNet model consists of fully convolutional network (FCN) layers arranged as contracting encoder and upsampling decoder maps. Nested arrangements of these encoder and decoder maps give rise to extensions of the UNet model, such as UNete and UNet++. Other refinements include constraining the outputs of the convolutional layers to discriminate between segment labels when trained end to end, a property called deep supervision. This reduces feature diversity in these nested UNet models despite their large parameter space. Furthermore, for texture segmentation, pixel correlations at multiple scales contribute to the classification task; hence, explicit deep supervision of shallower layers is likely to enhance performance. In this paper, we propose ADS UNet, a stage-wise additive training algorithm that incorporates resource-efficient deep supervision in shallower layers and takes performance-weighted combinations of the sub-UNets to create the segmentation model. We provide empirical evidence on three histopathology datasets to support the claim that the proposed ADS UNet reduces correlations between constituent features and improves performance while being more resource efficient. We demonstrate that ADS_UNet outperforms state-of-the-art Transformer-based models by 1.08 and 0.6 points on CRAG and BCSS datasets, and yet requires only 37% of GPU consumption and 34% of training time as that required by Transformers.

eess.IV↗

arXiv2023

Scale-Equivariant UNet for Histopathology Image Segmentation

Digital histopathology slides are scanned and viewed under different magnifications and stored as images at different resolutions. Convolutional Neural Networks (CNNs) trained on such images at a given scale fail to generalise to those at different scales. This inability is often addressed by augmenting training data with re-scaled images, allowing a model with sufficient capacity to learn the requisite patterns. Alternatively, designing CNN filters to be scale-equivariant frees up model capacity to learn discriminative features. In this paper, we propose the Scale-Equivariant UNet (SEUNet) for image segmentation by building on scale-space theory. The SEUNet contains groups of filters that are linear combinations of Gaussian basis filters, whose scale parameters are trainable but constrained to span disjoint scales through the layers of the network. Extensive experiments on a nuclei segmentation dataset and a tissue type segmentation dataset demonstrate that our method outperforms other approaches, with much fewer trainable parameters.

cs.CV↗

arXiv2023

Rotation-Scale Equivariant Steerable Filters

Incorporating either rotation equivariance or scale equivariance into CNNs has proved to be effective in improving models' generalization performance. However, jointly integrating rotation and scale equivariance into CNNs has not been widely explored. Digital histology imaging of biopsy tissue can be captured at arbitrary orientation and magnification and stored at different resolutions, resulting in cells appearing in different scales. When conventional CNNs are applied to histopathology image analysis, the generalization performance of models is limited because 1) a part of the parameters of filters are trained to fit rotation transformation, thus decreasing the capability of learning other discriminative features; 2) fixed-size filters trained on images at a given scale fail to generalize to those at different scales. To deal with these issues, we propose the Rotation-Scale Equivariant Steerable Filter (RSESF), which incorporates steerable filters and scale-space theory. The RSESF contains copies of filters that are linear combinations of Gaussian filters, whose direction is controlled by directional derivatives and whose scale parameters are trainable but constrained to span disjoint scales in successive layers of the network. Extensive experiments on two gland segmentation datasets demonstrate that our method outperforms other approaches, with much fewer trainable parameters and fewer GPU resources required. The source code is available at: https://github.com/ynulonger/RSESF.

cs.CV↗

arXiv2022

Re-evaluating Jet Reconstruction Techniques for New Higgs Boson Searches

The ultimate motivation of our study is to look for signs of physics beyond the Standard Model (BSM). We investigate whether different jet clustering techniques might be more or less suited to the particular final states of interest. In particular, we are interested in fully hadronic final states emerging from the decay chain of the Standard Model like Higgs boson into pairs of light Higgs states, the latter in turn decaying into bottom-anti bottom pairs. We show that, the ability of selecting the multi-jet final state and to reconstruct invariant masses of the Higgs bosons from it depend strongly on the choice of acceptance cuts, resolution parameters and reconstruction algorithm as well as its settings. Hence, we indicate the optimal choice of the latter for the purpose of establishing such a benchmark as a BSM signal. We then repeat the exercise for a heavy Higgs boson cascading into two SM-like Higgs states, obtaining similar results.

hep-ph↗

arXiv2022

Weight Fixing Networks

Modern iterations of deep learning models contain millions (billions) of unique parameters, each represented by a b-bit number. Popular attempts at compressing neural networks (such as pruning and quantisation) have shown that many of the parameters are superfluous, which we can remove (pruning) or express with less than b-bits (quantisation) without hindering performance. Here we look to go much further in minimising the information content of networks. Rather than a channel or layer-wise encoding, we look to lossless whole-network quantisation to minimise the entropy and number of unique parameters in a network. We propose a new method, which we call Weight Fixing Networks (WFN) that we design to realise four model outcome objectives: i) very few unique weights, ii) low-entropy weight encodings, iii) unique weight values which are amenable to energy-saving versions of hardware multiplication, and iv) lossless task-performance. Some of these goals are conflicting. To best balance these conflicts, we combine a few novel (and some well-trodden) tricks; a novel regularisation term, (i, ii) a view of clustering cost as relative distance change (i, ii, iv), and a focus on whole-network re-use of weights (i, iii). Our Imagenet experiments demonstrate lossless compression using 56x fewer unique weights and a 1.9x lower weight-space entropy than SOTA quantisation approaches.

cs.LG↗

arXiv2022

"Teaching Independent Parts Separately" (TIPSy-GAN) : Improving Accuracy and Stability in Unsupervised Adversarial 2D to 3D Pose Estimation

We present TIPSy-GAN, a new approach to improve the accuracy and stability in unsupervised adversarial 2D to 3D human pose estimation. In our work we demonstrate that the human kinematic skeleton should not be assumed as a single spatially codependent structure; in fact, we posit when a full 2D pose is provided during training, there is an inherent bias learned where the 3D coordinate of a keypoint is spatially codependent on the 2D coordinates of all other keypoints. To investigate our hypothesis we follow previous adversarial approaches but train two generators on spatially independent parts of the kinematic skeleton, the torso and the legs. We find that improving the self-consistency cycle is key to lowering the evaluation error and therefore introduce new consistency constraints during training. A TIPSy model is produced via knowledge distillation from these generators which can predict the 3D ordinates for the entire 2D pose with improved results. Furthermore, we address an unanswered question in prior work of how long to train in a truly unsupervised scenario. We show that for two independent generators training adversarially has improved stability than that of a solo generator which collapses. TIPSy decreases the average error by 17\% when compared to that of a baseline solo generator on the Human3.6M dataset. TIPSy improves upon other unsupervised approaches while also performing strongly against supervised and weakly-supervised approaches during evaluation on both the Human3.6M and MPI-INF-3DHP datasets.

cs.CV↗

arXiv2022

Optimising 2D Pose Representation: Improve Accuracy, Stability and Generalisability Within Unsupervised 2D-3D Human Pose Estimation

This paper addresses the problem of 2D pose representation during unsupervised 2D to 3D pose lifting to improve the accuracy, stability and generalisability of 3D human pose estimation (HPE) models. All unsupervised 2D-3D HPE approaches provide the entire 2D kinematic skeleton to a model during training. We argue that this is sub-optimal and disruptive as long-range correlations are induced between independent 2D key points and predicted 3D ordinates during training. To this end, we conduct the following study. With a maximum architecture capacity of 6 residual blocks, we evaluate the performance of 5 models which each represent a 2D pose differently during the adversarial unsupervised 2D-3D HPE process. Additionally, we show the correlations between 2D key points which are learned during the training process, highlighting the unintuitive correlations induced when an entire 2D pose is provided to a lifting model. Our results show that the most optimal representation of a 2D pose is that of two independent segments, the torso and legs, with no shared features between each lifting network. This approach decreased the average error by 20\% on the Human3.6M dataset when compared to a model with a near identical parameter count trained on the entire 2D kinematic skeleton. Furthermore, due to the complex nature of adversarial learning, we show how this representation can also improve convergence during training allowing for an optimum result to be obtained more often.

cs.CV↗

arXiv2022

Revisiting Jet Clustering Algorithms for New Higgs Boson Searches in Hadronic Final States

We assess the performance of different jet-clustering algorithms, in the presence of different resolution parameters and reconstruction procedures, in resolving fully hadronic final states emerging from the chain decay of the discovered Higgs boson into pairs of new identical Higgs states, the latter in turn decaying into bottom-antibottom quark pairs. We show that, at the Large Hadron Collider (LHC), both the efficiency of selecting the multi-jet final state and the ability to reconstruct from it the masses of the Higgs bosons (potentially) present in an event sample depend strongly on the choice of acceptance cuts, jet-clustering algorithm as well as its settings. Hence, we indicate the optimal choice of the latter for the purpose of establishing such a benchmark Beyond the SM (BSM) signal.

hep-ph↗