arXiv subjects

Supratik Mukhopadhyay

Publications and source records attributed to Supratik Mukhopadhyay.

At least 19 recordsLinked to original sources

arXiv2026

Experimental Characterization of Bulk Micromegas for Development of Active Target Time Projection Chamber for Nuclear Astrophysics Studies

A Micromegas-based active target Time Projection Chamber, namely Saha Active Target TPC (SAT-TPC), is under development at Saha Institute of Nuclear Physics for its application in nuclear astrophysics experiment to measure the branching ratio of the direct and sequential decay of the Hoyle state of $^{12}$C. A bulk Micromegas was characterized to optimize its operating parameters and its performance for $α$-particle tracking was investigated using gas mixtures Ar + CO$_{2}$ (90:10) and Ar + i-C$_4$H$_{10}$ (95:5) at atmospheric pressure using 5.9~keV X-ray and 5.48~MeV $α$-particles from radioactive sources. A hydrodynamic transport model was used to simulate the charge-deposit profile of the $α$-particle track on the Micromegas readout and compared to the experimental data to complement the experimental measurement as well as validate the numerical model. A good agreement was observed that confirmed the suitability of the choice of the bulk Micromegas in the development of the SAT-TPC prototype and the hydrodynamic modelling of the detector dynamics

physics.ins-det↗

arXiv2026

Beyond Static Costs: Learning-Dynamics Aware Loss Functions for Long-Tailed Classification

Deep learning models in computer vision face significant challenges when trained on long-tailed datasets, where a few majority classes dominate while many minority classes are severely underrepresented. Such imbalances frequently arise in real-world scenarios such as rare species recognition, manufacturing fault detection, and medical image understanding, leading to biased models that underperform on tail classes. Existing reweighting methods typically rely on static class frequencies to penalize the model, ignoring the dynamic nature of how effectively a network actually learns a class over time. We address this by introducing a novel Learning-Dynamics Aware Loss (LDAL) function that shifts the focus from static sample counts to dynamic learning progress. LDAL framework adjusts class weights continuously by leveraging: (i) the strength of learned feature representations (semantic scale), (ii) the intrinsic learning difficulty of each class, measured via the Shannon entropy of its predictions, and (iii) an inter-epoch regularizer term that tracks prediction shifts between consecutive epochs to stabilize training and avoid local minima. LDAL is purely a objective function which incurs negligible computational overhead while adapting to the feature learning of the model. Experimental results on multiple benchmark datasets demonstrate that our approach significantly surpasses state-of-the-art reweighting loss functions, providing an optimal trade-off between accuracy and generalizability. The source code is available at https://github.com/sdm2026/ldal

cs.CV↗

arXiv2026

neBEM: A GPU-accelerated Electrostatic Field Solver

Accurate electric field estimation is critical for the design and optimization of Micro Pattern Gaseous Detectors (MPGDs). The nearly exact Boundary Element Method (neBEM) offers high precision field computation but is limited by long CPU runtime arising from its complex analytical formulations. This work presents a comprehensive optimization of the neBEM solver, focusing on a hybrid hardware acceleration strategy using OpenMP for multi-core CPUs and GPU acceleration using NVIDIA's CUDA. A key contribution is the new implementation of a dynamic space charge calculation, which has also been designed to be accelerated by CUDA. This primary acceleration is complemented by enhanced algorithmic optimizations to reduce the complexity of the problem. The proposed implementation achieves substantial speedups while preserving inherent accuracy of the solver. Simulations on staggered thick Gas Electron Multiplier geometries demonstrate agreement with other commercially available field solvers, verifying the fidelity of accelerated neBEM. Benchmarking tests show a significant speedup, enabling rapid yet precise simulations for complex MPGD configurations. These improvements make GPU-accelerated neBEM a practical tool for large-scale detector simulation.

physics.ins-det↗

arXiv2026

Investigation of Thick-GEM detectors fabricated in India for muography application

Muography, commonly known as muon tomography, is a passive, non-destructive imaging technique that utilizes naturally occurring cosmic-ray muons to visualize the internal density structures of large, static, or inaccessible objects. In course of developing a small prototype muography system for material identification that relies upon the multiple Coulomb scattering of muons in matter, we explored the possible use of Thick-GEM detector as muon tracking device. It is a 5-20 fold scaled-up version of traditional GEM technology, that has become increasingly popular in recent years, owing to its mechanical robustness, cost-effective production, and excellent position sensing capabilities. A few prototypes of this detector of dimension $40\,\mathrm{mm} \times 48\,\mathrm{mm}$ with variation in other design parameters, were manufactured from a local industry. Subsequent to conditioning, detailed characterization of the detectors was performed to validate their suitability in muography applications. To identify the optimal operating region, gain variation was studied under various voltage configurations for both single and double-stage configurations. Experimental measurement of muon detection efficiency across the entire operating range yielded a maximum efficiency of 99.5\% in both cases. Using a collimated Fe$^{55}$-source, the best spatial resolution was determined to be 30 $μ$m for both single and double-stage operation.

physics.ins-det↗

arXiv2026

FPGA-Based Data Acquisition System for Muon Scattering Tomography

Muon scattering tomography is a non-destructive imaging technique that utilizes the penetrating properties and multiple Coulomb scattering of muons to produce detailed internal images of objects. This information is crucial for various applications, including material identification, civil structure investigation, geological surveys, archaeological investigations, and industrial inspections. In this work, we present the development of a multi-channel DAQ for muon tracking in a muon scattering tomography setup. The salient features of the proposed DAQ system include the direct acquisition and processing of LVDS signals, a 500MHz sampling frequency, and scalability. It consists of a front-end electronics stage built around the NINO ASIC, while the back-end electronics are configured with an Intel Altera MAX-10 FPGA development board, which transmits data to storage following the UART protocol. The performance of the DAQ system is tested with an RPC prototype to demonstrate its capability to accurately acquire two-dimensional position information of muon events. The proposed system provides an effective solution for large-scale muon tracking applications in non-destructive evaluation techniques such as muon scattering tomography.

physics.ins-det↗

arXiv2026

Benchmarking Artificial Intelligence Models for Daily Coastal Hypoxia Forecasting

Coastal hypoxia, especially in the northern part of Gulf of Mexico, presents a persistent ecological and economic concern. Seasonal models offer coarse forecasts that miss the fine-scale variability needed for daily, responsive ecosystem management. We present study that compares four deep learning architectures for daily hypoxia classification: Bidirectional Long Short-Term Memory (BiLSTM), Medformer (Medical Transformer), Spatio-Temporal Transformer (ST-Transformer), and Temporal Convolutional Network (TCN). We trained our models with twelve years of daily hindcast data from 2009-2020 Our training data consists of 2009-2020 hindcast data from a coupled hydrodynamic-biogeochemical model. Similarly, we use hindcast data from 2020 through 2024 as a test data. We constructed classification models incorporating water column stratification, sediment oxygen consumption, and temperature-dependent decomposition rates. We evaluated each architectures using the same data preprocessing, input/output formulation, and validation protocols. Each model achieved high classification accuracy and strong discriminative ability with ST-Transformer achieving the highest performance across all metrics and tests periods (AUC-ROC: 0.982-0.992). We also employed McNemar's method to identify statistically significant differences in model predictions. Our contribution is a reproducible framework for operational real-time hypoxia prediction that can support broader efforts in the environmental and ocean modeling systems community and in ecosystem resilience. The source code is available https://github.com/rmagesh148/hypoxia-ai/

cs.LG↗

arXiv2026

COMBOOD: A Semiparametric Approach for Detecting Out-of-distribution Data for Image Classification

Identifying out-of-distribution (OOD) data at inference time is crucial for many machine learning applications, especially for automation. We present a novel unsupervised semi-parametric framework COMBOOD for OOD detection with respect to image recognition. Our framework combines signals from two distance metrics, nearest-neighbor and Mahalanobis, to derive a confidence score for an inference point to be out-of-distribution. The former provides a non-parametric approach to OOD detection. The latter provides a parametric, simple, yet effective method for detecting OOD data points, especially, in the far OOD scenario, where the inference point is far apart from the training data set in the embedding space. However, its performance is not satisfactory in the near OOD scenarios that arise in practical situations. Our COMBOOD framework combines the two signals in a semi-parametric setting to provide a confidence score that is accurate both for the near-OOD and far-OOD scenarios. We show experimental results with the COMBOOD framework for different types of feature extraction strategies. We demonstrate experimentally that COMBOOD outperforms state-of-the-art OOD detection methods on the OpenOOD (both version 1 and most recent version 1.5) benchmark datasets (for both far-OOD and near-OOD) as well as on the documents dataset in terms of accuracy. On a majority of the benchmark datasets, the improvements in accuracy resulting from the COMBOOD framework are statistically significant. COMBOOD scales linearly with the size of the embedding space, making it ideal for many real-life applications.

cs.CV↗

arXiv2025

A Multimodal Human Protein Embeddings Database: DeepDrug Protein Embeddings Bank (DPEB)

Computationally predicting protein-protein interactions (PPIs) is challenging due to the lack of integrated, multimodal protein representations. DPEB is a curated collection of 22,043 human proteins that integrates four embedding types: structural (AlphaFold2), transformer-based sequence (BioEmbeddings), contextual amino acid patterns (ESM-2: Evolutionary Scale Modeling), and sequence-based n-gram statistics (ProtVec]). AlphaFold2 protein structures are available through public databases (e.g., AlphaFold2 Protein Structure Database), but the internal neural network embeddings are not. DPEB addresses this gap by providing AlphaFold2-derived embeddings for computational modeling. Our benchmark evaluations show GraphSAGE with BioEmbedding achieved the highest PPI prediction performance (87.37% AUROC, 79.16% accuracy). The framework also achieved 77.42% accuracy for enzyme classification and 86.04% accuracy for protein family classification. DPEB supports multiple graph neural network methods for PPI prediction, enabling applications in systems biology, drug target identification, pathway analysis, and disease mechanism studies.

cs.LG↗

arXiv2025

Classification of Hoyle State Decay Branches in Active Target Time Projection Chamber using Neural Network

A multi-class convolutional neural network (CNN) model has been developed using Keras deep learning library in Python for image-based classification of $^{12}$C Hoyle state decay branches from tracking information, recorded by Saha Active Target Time Projection Chamber, SAT-TPC (currently under development). The nuclear events, produced by the 30 MeV $α$-particle beam in the SAT-TPC, filled with Ar + CO$_2$ (90:10) gas mixture at atmospheric pressure, have been considered for training and validation of the models. The elastic scattering and Hoyle state sequential and direct decay events in the interaction of $α$-particle with $^{40}$Ar, $^{12}$C, $^{16}$O nuclei have been generated through Monte-Carlo simulation. The three-dimensional tracks, produced by the scattering and decay products through primary ionization of gaseous medium, have been simulated with Geant4. The primary tracks, distributed on the beam-plane, have been convoluted with electron diffusion, obtained with Magboltz, to produce the final tracking information. The classification performance of the proposed model for different readout segmentation schemes of the SAT-TPC has been discussed.

physics.ins-det↗

arXiv2025

Application of muon absorption tomography in imaging of civil structures

The research focuses on the non-invasive imaging technique using cosmic muon absorption tomography to monitor the internals of archaeological / civil / industrial structures of intermediate size. It integrates experimental measurements and numerical simulations with Geant4, ascertaining the reliability and precision of muon absorption tomography using easily available components for the stated purpose. The experiment probes muon interactions across a range of materials including those commonly used in building civil and industrial structures. An experiment, fondly named MARS (Muon Absorption in Rigid Structures), was carried out to explore the possibility of using overlapped scintillation paddles for improved mapping of inhomogeneities in structures made of concrete. Good correlation of experimental and simulated results for all tests indicates that this simple approach can be implemented for non-destructive evaluations of structures of civil and industrial interest

physics.ins-det↗

arXiv2024

Numerical Modelling of Active Target Time Projection Chamber for Low Energy Nuclear Physic

A numerical model based on hydrodynamic approach has been developed to emulate the device dynamics of active target Time Projection Chamber which is utilized for studying nuclear reaction through three dimensional tracking of concerned low energy particles. The proposed model has been used to investigate the performance of a prototype active target Time Projection Chamber, namely SAT-TPC, to be fabricated at Saha Institute of Nuclear Physics, for its application in nuclear physics experiments. A case study of non-relativistic elastic scattering $^4He+^{12}C$ with beam energy $25~MeV$ and current $2.3~pA$ has been opted for this purpose. The effect of beam induced space charge on the tracking performance the SAT-TPC prototype has been studied to optimize the beam current and scheme of the anode readout segmentation. The model has been validated by comparing its results to that of a particle model used to explain observed distortion in scattered particle tracks in a low energy nuclear physics experiment.

physics.ins-det↗

arXiv2024

Risks and Opportunities of Open-Source Generative AI

Applications of Generative AI (Gen AI) are expected to revolutionize a number of different areas, ranging from science & medicine to education. The potential for these seismic changes has triggered a lively debate about the potential risks of the technology, and resulted in calls for tighter regulation, in particular from some of the major tech companies who are leading in AI development. This regulation is likely to put at risk the budding field of open-source generative AI. Using a three-stage framework for Gen AI development (near, mid and long-term), we analyze the risks and opportunities of open-source generative AI models with similar capabilities to the ones currently available (near to mid-term) and with greater capabilities (long-term). We argue that, overall, the benefits of open-source Gen AI outweigh its risks. As such, we encourage the open sourcing of models, training and evaluation data, and provide a set of recommendations and best practices for managing risks associated with open-source generative AI.

cs.LG↗

arXiv2024

Near to Mid-term Risks and Opportunities of Open-Source Generative AI

In the next few years, applications of Generative AI are expected to revolutionize a number of different areas, ranging from science & medicine to education. The potential for these seismic changes has triggered a lively debate about potential risks and resulted in calls for tighter regulation, in particular from some of the major tech companies who are leading in AI development. This regulation is likely to put at risk the budding field of open-source Generative AI. We argue for the responsible open sourcing of generative AI models in the near and medium term. To set the stage, we first introduce an AI openness taxonomy system and apply it to 40 current large language models. We then outline differential benefits and risks of open versus closed source AI and present potential risk mitigation, ranging from best practices to calls for technical and scientific contributions. We hope that this report will add a much needed missing voice to the current public discourse on near to mid-term AI safety and other societal impact.

cs.LG↗

arXiv2024

Numerical simulation of charging up, accumulation of space charge and formation of discharges

Aging and stability of gaseous ionization detectors are intricately related to charging up, accumulation of space charge and formation of discharges. All these phenomena, in their turn, depend on the dynamics of charged particles within the device. Because of the large number of particles involved and their complex interactions, the dynamic processes of generation and loss of charged particles, and their transport within the detector volume are extremely expensive to simulate numerically. In this work, we propose and evaluate possible algorithms / approaches that show some promise in relation to the above-mentioned problems. Several important ionization detectors having parallel plate configurations, such as GEM, Micromegas, RPCs and THGEMs, are considered for this purpose. Information related to primary ionization is obtained from HEED, while all the transport properties are evaluated using MAGBOLTZ. The transport dynamics have been followed using two different approaches. In one, particle description using neBEM-Garfield++ combination has been used. For this purpose, the neBEM solver has been significantly improved such that perturbations due to the charged particles present within the device are considered while estimating electric field. In the other approach, the transport is simulated following hydrodynamic model using COMSOL during which the electric field is also provided by COMSOL where it is easy to set up space charge effects. A comparison between these possible approaches will be presented. Effect of different simulation parameters will also be demonstrated using simple examples.

physics.ins-det↗

arXiv2024

Effect of hole geometry on charge sharing and other parameters in GEM-based detectors

Gas Electron Multipliers (GEM) are among the more prominent Micro-Pattern Gaseous Detectors (MPGDs) and widely used in high energy particle physics experiments and various related applications. Adoption of different production techniques lead to holes of varying geometries in GEM foils. Since the response of a GEM-based detector is closely related to the hole geometry through the influence of the latter on charge sharing and transport through GEM foils, attempts have been made to relate hole configurations to different figures of merit of a detector. Numerical simulations have been performed to study the effects of hole geometry on important parameters such as charge sharing, collection efficiency, extraction efficiency, gain, possibility of transition from avalanche to streamer modes for single, double and triple layer GEM detectors. The numerical estimates have been compared to available experimental data. The comparisons, although not always in agreement, are found to be generally encouraging.

physics.ins-det↗

arXiv2023

Parallelization of Garfield++ and neBEM to simulate space charge effects in RPCs

Numerical simulation of avalanches, saturated avalanches, and streamers can help us understand the detector physics of Resistive Plate Chambers (RPC). 3D Monte Carlo simulation of an avalanche inside an RPC, the transition from avalanche to saturated avalanche to streamer may help the search for the optimum voltage and alternate gas mixtures. This task is dauntingly resource-hungry, especially when space charge effects become important, which often coincides with important regimes of operation of these devices. By modifying the electric field inside the RPC dynamically, the space charge plays a crucial role in determining the response of the detector. In this work, a numerical model has been proposed to calculate the dynamic space-charge field inside an RPC and the same has been implemented in the Garfield++ framework. By modeling space charge as the large number of line charges and using the multithreading technique OpenMP to calculate electric field, drift line, electron gain, and space charge field, it has been possible to maintain time consumption within reasonable limits. For this purpose, a new class, pAvalancheMC has been introduced in Garfield++. The calculations have been successfully verified with those from existing solvers and an example is provided to show the performance of pAvalancheMC. Moreover, the details of the transition of an avalanche into a saturated avalanche have been discussed. The induced charge distribution is calculated for a timing RPC and results are verified with the experiment.

physics.ins-det↗

arXiv2023

Determination of charge spread, position resolution, energy resolution and gain uniformity of Gas Electron Multipliers (GEM)

Gas electron multipliers (GEM) detectors are gaseous detectors widely used for tracking and imaging applications due to their good position resolution, high efficiency at high irradiation rates, among other factors. In the present work, position resolution, charge spread, energy resolution and gain uniformity have been investigated experimentally for single and double GEM geometries using an Fe-55 source. The position resolution measurements have been performed by a novel method, using a high precision instrument for source movement and is found to be highly successful. The result shows that the double GEM can resolve positions with sigma values up to 36.8 micron and 54.6 micron in x and y directions, respectively. To validate the experimental results, a Garfield simulation work has been carried out on charge spread.

physics.ins-det↗

arXiv2022

Design and studies of thick Gas Electron Multipliers fabricated in India

THick Gas Electron Multipliers (THGEMs) are robust and high gain Micro Pattern Gaseous Detectors which are economically manufactured by standard drilling and etching of thin printed circuit boards. In this paper, we present our recent simulation as well as experimental studies on THGEMs which have been fabricated in India using local expertise. Two types of THGEMs have been fabricated; one set has holes without any external rim and another set has holes with rims. These detectors have been characterized using argon-carbon dioxide and argon-isobutane gas mixtures. Electron transmission, effective gain, energy resolution and optimized working range studies have been presented for both the sets of THGEMs.

physics.ins-det↗