Faculty Publications
Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736
Publications by NITK Faculty
Browse
8 results
Search Results
Item Dynamic pricing of call rates: Bayesian approach(Elsevier B.V., 2015) Dugar, C.; Jain, A.; Rajawat, A.; Bhattacharya, S.In this paper, we present different cases and their possible solutions in the telecommunications market by incorporating dynamically changing call rates over the channel depending upon the network congestion. Since dynamic pricing of call rates is beneficial from both the perspectives of subscribers and service providers, our solution can significantly help to adapt this pricing mechanism in real market scenario. In order to deploy this scheme, we have incorporated the competing network provider's strategy into the mechanism of deciding dynamic price. Establishment of Nash equilibrium with the competing network provider has stabilized our pricing mechanism. © 2014 Elsevier B.V. All rights reserved.Item Fault diagnosis studies of face milling cutter using machine learning approach(Multi-Science Publishing Co. Ltd claims@sagepub.com, 2016) Madhusudana, C.K.; Budati, S.; Gangadhar, N.; Kumar, H.; Narendranath, S.Successful automation of a machining process system requires an effective and efficient tool condition monitoring system to ensure high productivity, products of desired dimensions, and long machine tool life. As such the component's processing quality and increased system reliability will be guaranteed. This paper presents a classification of healthy and faulty conditions of the face milling tool by using the Naive Bayes technique. A set of descriptive statistical parameters is extracted from the vibration signals. The decision tree technique is used to select significant features out of all statistical extracted features. The selected features are fed to the Naive Bayes algorithm. The output of the algorithm is used to study and classify the milling tool condition and it is found that the Naive Bayes model is able to give 96.9% classification accuracy. Also the performances of the different classifiers are compared. Based on the results obtained, the Naive Bayes technique can be recommended for online monitoring and fault diagnosis of the face milling tool. © 2016 The Author(s).Item A Bayesian inference approach: estimation of heat flux from fin for perturbed temperature data(Springer India, 2018) Kumar, H.; Gnanasekaran, N.This paper reports the estimation of the unknown boundary heat flux from a fin using the Bayesian inference method. The setup consists of a rectangular mild steel fin of dimensions 250×150×6 mm3 and an aluminium base plate of dimensions 250×150×8 mm3. The fin is subjected to constant heat flux at the base and the fin setup is modelled using ANSYS14.5. The problem considered is a conjugate heat transfer from the fin, and the Navier–Stokes equation is solved to obtain the flow parameters. Grid independence study is carried out to fix the number of grids for the study considered. To reduce the computational cost, computational fluid dynamics (CFD) is replaced with artificial neural network (ANN) as the forward model. The Markov Chain Monte Carlo (MCMC) powered by Metropolis–Hastings sampling algorithm along with the Bayesian framework is used to explore the estimation space. The sensitivity analysis of the estimated temperature with respect to the unknown parameter is discussed to know the dependency of the temperature with the parameter. This paper signifies the effect of a prior model on the execution of the inverse algorithm at different noise levels. The unknown heat flux is estimated for the surrogated temperature and the estimates are reported as mean, Maximum a Posteriori (MAP) and standard deviation. The effect of a-priori information on the estimated parameter is also addressed. The standard deviation in the estimation process is referred to as the uncertainty associated with the estimated parameters. © 2018, Indian Academy of Sciences.Item Simultaneous estimation of unknown parameters using a-priori knowledge for the estimation of interfacial heat transfer coefficient during solidification of Sn–5wt%Pb alloy—an ANN-driven Bayesian approach(Springer, 2019) Vishweshwara, P.S.; Gnanasekaran, N.; Arun, M.The present methodology focuses on model reduction in which the prevalent one-dimensional transient heat conduction equation for a horizontal solidification of Sn–5wt%Pb alloy is replaced with Artificial Neural Network (ANN) in order to estimate the unknown constants present in the interfacial heat transfer coefficient correlation. As a novel approach, ANN-driven forward model is synergistically combined with Bayesian framework and Genetic algorithm to simultaneously estimate the unknown parameters and modelling error. Gaussian noise is then added to the temperature distribution obtained using the forward approach to represent real-time experiments. The hallmark of the present work is to reduce the computational time of both the forward and the inverse methods and to simultaneously estimate the unknown parameters using a-priori engineering knowledge. The results of the present methodology prove that the simultaneous estimation of unknown parameters can be effectively obtained only with the use of Bayesian framework. © 2019, Indian Academy of Sciences.Item MICAnet: A Deep Convolutional Neural Network for mineral identification on Martian surface(Elsevier B.V., 2024) Kumari, P.; Soor, S.; Shetty, A.; Koolagudi, S.G.Mineral identification plays a vital role in understanding the diversity and past habitability of the Martian surface. Mineral mapping by the traditional manual method is time-consuming and the unavailability of ground truth data limited the research on building supervised learning models. To address this issue an augmentation process is already proposed in the literature that generates training data replicating the spectra in the MICA (Minerals Identified in CRISM Analysis) spectral library while preserving absorption signatures and introducing variability. This study introduces MICAnet, a specialized Deep Convolutional Neural Network (DCNN) architecture for mineral identification using the CRISM (Compact Reconnaissance Imaging Spectrometer for Mars) hyperspectral data. MICAnet is inspired by the Inception-v3 and InceptionResNet-v1 architectures, but it is tailored with 1-dimensional convolutions for processing the spectra at the pixel level of a hyperspectral image. To the best of the authors’ knowledge, this is the first DCNN architecture solely dedicated to mineral identification on the Martian surface. The model is evaluated by its matching with a TRDR (Targeted Reduced Data Record) dataset obtained using a hierarchical Bayesian model. The results demonstrate an impressive f-score of at least .77 among different mineral groups in the MICA library, which is on par with or better than the unsupervised models previously applied to this objective. © 2024Item Optimizing Hyperparameters in Meta-Learning for Enhanced Image Classification(Institute of Electrical and Electronics Engineers Inc., 2025) Vincent, A.M.; Padikkal, P.; Bini, A.A.This paper investigates the significance of hyperparameter optimization in meta-learning for image classification tasks. Despite advancements in deep learning, real-time image classification applications often suffer from data inadequacy. Few-shot learning addresses this challenge by enabling learning from limited samples. Meta-learning, a prominent tool for few-shot learning, learns across multiple classification tasks. We explore different types of meta-learners, with a particular focus on metric-based models. We analyze the potential of hyperparameter optimization techniques, specifically Bayesian optimization and its variants, to enhance the performance of these models. Experimental results on the Omniglot and ImageNet datasets demonstrate that incorporating Bayesian optimization, particularly its evolutionary strategy variant, into meta-learning frameworks leads to improved accuracy compared to settings without hyperparameter optimization. Here, we show that by optimizing hyperparameters for individual tasks rather than using a uniform setting, we achieve notable gains in model performance, underscoring the importance of tailored hyperparameter configurations in meta-learning. © 2013 IEEE.Item Comparison of the Multiple Imputation Approaches for Imputing Rainfall Data: A Humid Tropical River Basin Case Study(Springer Nature, 2025) Kumar, G.P.; Dwarakish, G.S.Accurate rainfall data is crucial for agriculture, hydrology, and climate research as it guides water management, crop planning, and disaster preparedness. Missing data affects reliability, requiring effective imputation. The purpose of this study is to address the critical challenge of imputing missing daily rainfall data, which is especially important given rainfall’s nonlinear distribution and variability in missingness patterns. This research aims to develop and evaluate advanced imputation algorithms to improve data completeness and integrity in humid tropical regions. This study evaluates ten imputation algorithms: K-nearest neighbors (KNN), classification and regression trees (CART), predictive mean matching (pmm), random forest (rf), mean method, and Bayesian methods (norm. boot, lasso. norm, norm, norm. nob, midastouch) for addressing missing daily rainfall data. Using 37 years of data from thirteen stations in the Kali River Basin, the methods leverage descriptive statistics to enhance accuracy in humid tropical regions. The proposed algorithms incorporate descriptive statistics of the rainfall time series and are evaluated using 37 years of daily data from thirteen selected rainfall stations in the Kali River Basin, a humid tropical region. Model performance was assessed at four missingness levels (1%, 5%, 10%, and 20%) and evaluated with accuracy metrics root mean square error (RMSE), mean absolute error (MAE), index of agreement (d), and RMSE-observations standard deviation ratio (RSR). Among the evaluated methods, KNN consistently demonstrated superior performance across all levels of missingness (RMSE = 13.22 to 15.42; MAE = 4.68 to 6.08; d = 0.87 to 0.90; RSR = 0.57 to 0.61), followed closely by CART (RMSE = 16.48 to 20.77; MAE = 6.20 to 8.31; d = 0.81 to 0.86; RSR = 0.71 to 0.79). Overall, KNN, CART, pmm, and rf emerged as reliable methods for imputing missing rainfall data of varying lengths, contributing to more accurate weather forecasting and climate change analyses. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.Item Enhancing Money Laundering Detection in Bank Transactions using GAGAN: A Graph-Adapted Generative Adversarial Network Approach(Springer Science and Business Media Deutschland GmbH, 2025) Kadamathikuttiyil Karthikeyan, G.; Bhowmik, B.The past decade has witnessed profound transformations in the financial sector, driven by the integration of cutting-edge technologies into its core operations. Consequently, banks are increasingly utilizing technologies such as artificial intelligence (AI), blockchain, and big data analytics to offer personalized services, streamline transactions, and improve risk management, enabling the development of new financial products and services that cater to the diverse and evolving needs of customers. Despite these benefits, the banking landscape has also brought about complex challenges, particularly in the fight against money laundering. Money laundering remains a significant threat to the integrity of financial systems, as criminals exploit digital advancements to conceal illicit activities. The growing complexity of digital transactions and the increasing volume of financial data have made detecting and preventing money laundering more challenging than ever. Existing AI-based solutions, while effective to some extent, often grapple with class imbalance issues. This paper addresses the challenge by introducing a novel model named GAGAN (Graph Attention Generative Adversarial Network) and enhances the detection of money laundering activities in bank transactions. The proposed model further addresses the issue of class imbalance, by incorporating Conditional Generative Adversarial Network (cGAN) and Graph Attention Networks (GAT). The GAT classifier is then employed to accurately classify transactions, leveraging attention mechanisms to focus on the most relevant parts of the graph. Empirical results reveal that the proposed model achieves impressive performance metrics, with an accuracy of 98.62%, precision of 98.10%, recall of 98.92%, F1 score of 98.49%, AUC-ROC of 0.99, and a MCC score of 0.991. These results underscore the efficacy of the model in accurately identifying money laundering transactions, showcasing its potential as a robust tool for financial crime detection. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.