arXiv subjects

Sourav Saha

Publications and source records attributed to Sourav Saha.

34 records · Page 2Linked to original sources

arXiv2025

Explainable Hierarchical Deep Learning Neural Networks (Ex-HiDeNN)

Data-driven science and computation have advanced immensely to construct complex functional relationships using trainable parameters. However, efficiently discovering interpretable and accurate closed-form expressions from complex dataset remains a challenge. The article presents a novel approach called Explainable Hierarchical Deep Learning Neural Networks or Ex-HiDeNN that uses an accurate, frugal, fast, separable, and scalable neural architecture with symbolic regression to discover closed-form expressions from limited observation. The article presents the two-step Ex-HiDeNN algorithm with a separability checker embedded in it. The accuracy and efficiency of Ex-HiDeNN are tested on several benchmark problems, including discerning a dynamical system from data, and the outcomes are reported. Ex-HiDeNN generally shows outstanding approximation capability in these benchmarks, producing orders of magnitude smaller errors compared to reference data and traditional symbolic regression. Later, Ex-HiDeNN is applied to three engineering applications: a) discovering a closed-form fatigue equation, b) identification of hardness from micro-indentation test data, and c) discovering the expression for the yield surface with data. In every case, Ex-HiDeNN outperformed the reference methods used in the literature. The proposed method is built upon the foundation and published works of the authors on Hierarchical Deep Learning Neural Network (HiDeNN) and Convolutional HiDeNN. The article also provides a clear idea about the current limitations and future extensions of Ex-HiDeNN.

cs.LG↗

arXiv2025

KHRONOS: a Kernel-Based Neural Architecture for Rapid, Resource-Efficient Scientific Computation

Contemporary models of high dimensional physical systems are constrained by the curse of dimensionality and a reliance on dense data. We introduce KHRONOS (Kernel Expansion Hierarchy for Reduced Order, Neural Optimized Surrogates), an AI framework for model based, model free and model inversion tasks. KHRONOS constructs continuously differentiable target fields with a hierarchical composition of per-dimension kernel expansions, which are tensorized into modes and then superposed. We evaluate KHRONOS on a canonical 2D, Poisson equation benchmark: across 16 to 512 degrees of freedom (DoFs), it obtained L_2-square errors of 5e-4 down to 6e-11. This represents a greater than 100-fold gain over Kolmogorov Arnold Networks (which itself reports a 100 times improvement on MLPs/PINNs with 100 times fewer parameters) when controlling for the number of parameters. This also represents a 1e6-fold improvement in L_2-square error compared to standard linear FEM at comparable DoFs. Inference complexity is dominated by inner products, yielding sub-millisecond full-field predictions that scale to an arbitrary resolution. For inverse problems, KHRONOS facilitates rapid, iterative level set recovery in only a few forward evaluations, with sub-microsecond per sample latency. KHRONOS's scalability, expressivity, and interpretability open new avenues in constrained edge computing, online control, computer vision, and beyond.

cs.LG↗

arXiv2025

INN-FF: A Scalable and Efficient Machine Learning Potential for Molecular Dynamics

The ability to accurately model interatomic interactions in large-scale systems is fundamental to understanding a wide range of physical and chemical phenomena, from drug-protein binding to the behavior of next-generation materials. While machine learning interatomic potentials (MLIPs) have made it possible to achieve ab initio-level accuracy at significantly reduced computational cost, they still require very large training datasets and incur substantial training time and expense. In this work, we propose the Interpolating Neural Network Force Field (INN-FF), a novel framework that merges interpolation theory and tensor decomposition with neural network architectures to efficiently construct molecular dynamics potentials from limited quantum mechanical data. Interpolating Neural Networks (INNs) achieve comparable or better accuracy than traditional multilayer perceptrons (MLPs) while requiring orders of magnitude fewer trainable parameters. On benchmark datasets such as liquid water and rMD17, INN-FF not only matches but often surpasses state-of-the-art accuracy by an order of magnitude, while achieving significantly lower error when trained on smaller datasets. These results suggest that INN-FF offers a promising path toward building efficient and scalable machine-learned force fields.

cond-mat.mtrl-sci↗

arXiv2025

ir_explain: a Python Library of Explainable IR Methods

While recent advancements in Neural Ranking Models have resulted in significant improvements over traditional statistical retrieval models, it is generally acknowledged that the use of large neural architectures and the application of complex language models in Information Retrieval (IR) have reduced the transparency of retrieval methods. Consequently, Explainability and Interpretability have emerged as important research topics in IR. Several axiomatic and post-hoc explanation methods, as well as approaches that attempt to be interpretable-by-design, have been proposed. This article presents \irexplain, an open-source Python library that implements a variety of well-known techniques for Explainable IR (ExIR) within a common, extensible framework. \irexplain supports the three standard categories of post-hoc explanations, namely pointwise, pairwise, and listwise explanations. The library is designed to make it easy to reproduce state-of-the-art ExIR baselines on standard test collections, as well as to explore new approaches to explaining IR models and methods. To facilitate adoption, \irexplain is well-integrated with widely-used toolkits such as Pyserini and \irdatasets.

cs.IR↗

arXiv2025

Combining Query Performance Predictors: A Reproducibility Study

A large number of approaches to Query Performance Prediction (QPP) have been proposed over the last two decades. As early as 2009, Hauff et al. [28] explored whether different QPP methods may be combined to improve prediction quality. Since then, significant research has been done both on QPP approaches, as well as their evaluation. This study revisits Hauff et al.s work to assess the reproducibility of their findings in the light of new prediction methods, evaluation metrics, and datasets. We expand the scope of the earlier investigation by: (i) considering post-retrieval methods, including supervised neural techniques (only pre-retrieval techniques were studied in [28]); (ii) using sMARE for evaluation, in addition to the traditional correlation coefficients and RMSE; and (iii) experimenting with additional datasets (Clueweb09B and TREC DL). Our results largely support previous claims, but we also present several interesting findings. We interpret these findings by taking a more nuanced look at the correlation between QPP methods, examining whether they capture diverse information or rely on overlapping factors.

cs.IR↗

arXiv2024

Interpolating neural network: A novel unification of machine learning and interpolation theory

Artificial intelligence (AI) has revolutionized software development, shifting from task-specific codes (Software 1.0) to neural network-based approaches (Software 2.0). However, applying this transition in engineering software presents challenges, including low surrogate model accuracy, the curse of dimensionality in inverse design, and rising complexity in physical simulations. We introduce an interpolating neural network (INN), grounded in interpolation theory and tensor decomposition, to realize Engineering Software 2.0 by advancing data training, partial differential equation solving, and parameter calibration. INN offers orders of magnitude fewer trainable/solvable parameters for comparable model accuracy than traditional multi-layer perceptron (MLP) or physics-informed neural networks (PINN). Demonstrated in metal additive manufacturing, INN rapidly constructs an accurate surrogate model of Laser Powder Bed Fusion (L-PBF) heat transfer simulation, achieving sub-10-micrometer resolution for a 10 mm path in under 15 minutes on a single GPU. This makes a transformative step forward across all domains essential to engineering software.

cs.LG↗

arXiv2023

A Comparative Analysis of CNN-Based Pretrained Models for the Detection and Prediction of Monkeypox

Monkeypox is a rare disease that raised concern among medical specialists following the convi-19 pandemic. It's concerning since monkeypox is difficult to diagnose early on because of symptoms that are similar to chickenpox and measles. Furthermore, because this is a rare condition, there is a knowledge gap among healthcare professionals. As a result, there is an urgent need for a novel technique to combat and anticipate the disease in the early phases of individual virus infection. Multiple CNN-based pre-trained models, including VGG-16, VGG-19, Restnet50, Inception-V3, Densnet, Xception, MobileNetV2, Alexnet, Lenet, and majority Voting, were employed in classification in this study. For this study, multiple data sets were combined, such as monkeypox vs chickenpox, monkeypox versus measles, monkeypox versus normal, and monkeypox versus all diseases. Majority voting performed 97% in monkeypox vs chickenpox, Xception achieved 79% in monkeypox against measles, MobileNetV2 scored 96% in monkeypox vs normal, and Lenet performed 80% in monkeypox versus all.

cs.CV↗

arXiv2022

An Introduction to Kernel and Operator Learning Methods for Homogenization by Self-consistent Clustering Analysis

Recent advances in operator learning theory have improved our knowledge about learning maps between infinite dimensional spaces. However, for large-scale engineering problems such as concurrent multiscale simulation for mechanical properties, the training cost for the current operator learning methods is very high. The article presents a thorough analysis on the mathematical underpinnings of the operator learning paradigm and proposes a kernel learning method that maps between function spaces. We first provide a survey of modern kernel and operator learning theory, as well as discuss recent results and open problems. From there, the article presents an algorithm to how we can analytically approximate the piecewise constant functions on R for operator learning. This implies the potential feasibility of success of neural operators on clustered functions. Finally, a k-means clustered domain on the basis of a mechanistic response is considered and the Lippmann-Schwinger equation for micro-mechanical homogenization is solved. The article briefly discusses the mathematics of previous kernel learning methods and some preliminary results with those methods. The proposed kernel operator learning method uses graph kernel networks to come up with a mechanistic reduced order method for multiscale homogenization.

cs.LG↗

arXiv2022

Deep Learning Discrete Calculus (DLDC): A Family of Discrete Numerical Methods by Universal Approximation for STEM Education to Frontier Research

The article proposes formulating and codifying a set of applied numerical methods, coined as Deep Learning Discrete Calculus (DLDC), that uses the knowledge from discrete numerical methods to interpret the deep learning algorithms through the lens of applied mathematics. The DLDC methods aim to leverage the flexibility and ever increasing resources of deep learning and rich literature of numerical analysis to formulate a general class of numerical method that can directly use data with uncertainty to predict the behavior of an unknown system as well as elevate the speed and accuracy of numerical solution of the governing equations for known systems. The article is structured in two major sections. In the first section, the building blocks of the DLDC methods are presented and deep learning structures analogous to traditional numerical methods such as finite difference and finite element methods are constructed with a view to incorporate these techniques in Science, Technology, Engineering, Mathematics (STEM) syllabus for K-12 students. The second section builds upon the building blocks of the previous discussion,and proposes new solution schemes for differential and integral equations pertinent to multiscale mechanics. Each section is accompanied with mathematical formulation of the numerical methods, analogous DLDC formulation, and suitable examples.

math.NA↗

arXiv2021

MEAM parameterization for cyclic and tensile deformations of Gold-Silver core-shell systems

Gold-Silver (Au-Ag) core-shell nanostructures are gaining importance in stretchable electronics where high tensile and fatigue resistance is of paramount importance. This work proposes the parameterization of a modified embedded atomic model (MEAM) interatomic potential through density functional theory (DFT) calculations for investigating the role of dislocations and defect interaction governing the mechanical behavior of Au-Ag and Ag-Au Core-shell nanostructures under tensile and fatigue loading using molecular dynamics (MD) simulations. A comparative analysis between the Core-shell structures and their pristine counterparts is also conducted. Throughout this work, pseudo-potential and all-electron full potential DFT schemes are used for parameterizing MEAM by calculating cohesive energy, lattice parameter, and bulk modulus of pure Au, Ag and their alloy. Using the new force-field for MD simulations, the tensile behavior of pristine and core-shell nanowires is explored for temperatures between 300K to 600K. The fatigue properties of two pristine and two core-shell nanowires in a strain range of -15% to 15% for 10 cycles is also conducted. Our results suggest that Ag-Au Core-shell nanowire shows the best reversibility under fatigue loading among the structures examined. Moreover, Ag-Au exhibit the highest dislocation formation and complete annihilation of defects consistently. While, Au-Ag present improved fatigue properties than its pristine counterparts but have some residual defects leading to lower reversibility when compared to Ag-Au. For tensile loading, all four structures exhibited deterioration in strength with increasing temperature. Thermal softening is seen to be more prominent in Au-Ag core-shell nanowires compared to Ag-Au.

cond-mat.mtrl-sci↗

arXiv2021

Bangla Image Caption Generation through CNN-Transformer based Encoder-Decoder Network

Automatic Image Captioning is the never-ending effort of creating syntactically and validating the accuracy of textual descriptions of an image in natural language with context. The encoder-decoder structure used throughout existing Bengali Image Captioning (BIC) research utilized abstract image feature vectors as the encoder's input. We propose a novel transformer-based architecture with an attention mechanism with a pre-trained ResNet-101 model image encoder for feature extraction from images. Experiments demonstrate that the language decoder in our technique captures fine-grained information in the caption and, then paired with image features, produces accurate and diverse captions on the BanglaLekhaImageCaptions dataset. Our approach outperforms all existing Bengali Image Captioning work and sets a new benchmark by scoring 0.694 on BLEU-1, 0.630 on BLEU-2, 0.582 on BLEU-3, and 0.337 on METEOR.

cs.CV↗

arXiv2020

Deformation mechanisms of Inconel-718 at nanoscale by molecular dynamics

Ni-based superalloy Inconel-718 is ubiquitous in metal 3D printing where high cooling rate and thermal gradient are present. These manufacturing conditions are conducive to high initial dislocation density and porosity or void in the material. This work proposes a molecular dynamics (MD) analysis method that can examine the role of dislocations, cooling rates, void, and their interactions governing the material properties and failure mechanism in Inconel-718 using Embedded Atom Method (EAM) potential. Three different structures: nanowire (NW), nanopillar, and nanoplate are used throughout this work. Initially, strain rates are varied from 10^8s^-1 to 10^10s^-1 keeping the NW diameter and temperature constant at 3.17 nm and 300K respectively. Compressive loading is applied to a 7.04 nm nanopillar by applying a constant strain rate of 109 s^-1 while temperature is varied from 100K to 700K. Different cooling rates ranging from 0.5x10^10 K/s to 1 x 10^14 K/s are applied to nanoplates (with and without a central void). The size of the central void is kept fixed at 2.12 nm. Increasing strain rates in tension not only results in strain hardening but also increase in dislocation density. Our computational method is successful to capture extensive sliding on {111} shear plane, which leads to significant necking of the alloy before fracture due to dislocation. The high cooling rates creating non-equilibrium structure leads to high strength and ductile behavior. On the other hand, the low cooling rate forming well defined crystalline structure causes low strength and brittle behavior. This brittle to ductile transition is observed solely due to cooling rate. Cooling rate may diminish the void by healing the structure during solidifications process. Subsequent mechanical properties by varying temperature and size are also presented in detail.

cond-mat.mtrl-sci↗

arXiv2019

Insights into the mechanical properties and fracture mechanism of Cadmium Telluride nanowire

The mechanical properties of Cadmium telluride (CdTe) nanowire have become focus of interest now-a-days due to its promising application in opto-electro-mechanical nanodevices. Inthis study, molecular dynamics simulations have been used to investigate the mechanical behavior of Zinc Blende (ZB) crystal structured CdTe nanowires (NWs) by varying size, temperature, crystal orientation and strain rate under tension and compression. Results show that the fracture strength of the [111]-oriented CdTe NWs is always higher than that of the [110]-oriented CdTe NWs under tension whereas in compression, the fracture strength of the [111]-oriented CdTe NWs is significantly lower than that of the [110]-oriented CdTe NWs. Moreover, under tensile load, void in ZB [111]-oriented CdTe NWs has been observed which is a new failure mechanism found in this study. It has also been observed that size has negligible effect on the tensile behavior but in compression the behavior is clearly size dependent. Both tensile and compressive strengths show an inverse relation with temperature. When tensile load is applied along NWs growth direction, the [111]-oriented CdTe NWs fail by creating void in [10-1] direction regardless of temperature and NW size. Under compression, the [111]-oriented nanowire show buckling and plasticity. Finally, the impact of strain rate on [111]-oriented ZB CdTe NWs is also studied where higher fracture strengths and strains at a higher strain rates have been found under both tension and compression. With increasing the strain rate, the number of voids is also increased in the NWs.This study will help to design CdTe NWs based devices efficiently by presenting in-depth understanding of failure behavior of the [111]-oriented CdTe NWs.

cond-mat.mtrl-sci↗

arXiv2017

Study of Uniaxial Tensile Properties of Hexagonal Boron Nitride Nanoribbons

Uniaxial tensile properties of hexagonal boron nitride nanoribbons and dependence of these properties on temperature, strain rate, and the inclusion of vacancy defects have been explored with molecular dynamics simulations using Tersoff potential. The ultimate tensile strength of pristine hexagonal boron nitride nanoribbon of 26 nm x 5 nm with armchair chirality is found to be 100.5 GPa. The ultimate tensile strength and strain have been found decreasing with increasing the temperature while an opposite trend has been observed for increasing the strain rate. Furthermore, the vacancy defects reduce ultimate tensile strength and strain where the effect of bi-vacancy is clearly dominating over point vacancy.

cond-mat.mtrl-sci↗

arXiv2017

Atomistic Representation of Anomalies in the Failure Behaviour of Nanocrystalline Silicene

Silicene, a 2D analogue of graphene, has spurred a tremendous research interest in the scientific community for its unique properties essential for next generation electronic devices. In this work, for the first time, we present a molecular dynamics (MD) investigation to determine the fracture strength and toughness of nanocrystalline silicene (nc silicene) sheet of varied grain size and pre existing crack length at room temperature. Our results suggest that the transition from an inverse pseudo Hall Petch to a pseudo Hall Petch behavior in nc silicene occurs at a critical grain size of 17.32 nm. This phenomenon is also prevalent in nanocrystalline graphene. However, nc silicene with pre existing cracks exhibits anomalous crack propagation and fracture toughness behaviour. We have observed two distinct types of failure mechanisms (crack sensitive and insensitive failure) and devised the mechanophysical conditions under which they occur. Fracture toughness calculated from both Griffiths theory and MD simulations indicate that the former overpredicts the fracture toughness of nc silicene. The most striking outcome, however, is that despite the presence of a pre existing crack, the crack sensitivity of nc silicene is found to be dependent on the grain size and their orientations. This study is the first direct comparison of atomistic simulations to the continuum theories to predict the anomalous behaviour in deformation and failure mechanisms of nc silicene.

cond-mat.mtrl-sci↗

arXiv2017

Graphene and its elemental analogue: A molecular dynamics view of fracture phenomenon

Graphene and some graphene like two dimensional materials; hexagonal boron nitride (hBN) and silicene have unique mechanical properties which severely limit the suitability of conventional theories used for common brittle and ductile materials to predict the fracture response of these materials. This study revealed the fracture response of graphene, hBN and silicene nanosheets under different tiny crack lengths by molecular dynamics (MD) simulations using LAMMPS. The useful strength of these large area two dimensional materials are determined by their fracture toughness. Our study shows a comparative analysis of mechanical properties among the elemental analogues of graphene and suggested that hBN can be a good substitute for graphene in terms of mechanical properties. We have also found that the pre-cracked sheets fail in brittle manner and their failure is governed by the strength of the atomic bonds at the crack tip. The MD prediction of fracture toughness shows significant difference with the fracture toughness determined by Griffth's theory of brittle failure which restricts the applicability of Griffith's criterion for these materials in case of nano-cracks. Moreover, the strengths measured in armchair and zigzag directions of nanosheets of these materials implied that the bonds in armchair direction has the stronger capability to resist crack propagation compared to zigzag direction.

cond-mat.mtrl-sci↗