Muthumanickam Gunasekaran

Machine learning models, especially vision transformers in the domain of medical images, are highly prone to data poisoning attacks, in which a small proportion of adversarial samples is injected into the model's training dataset to manipulate its behavior. Existing data poisoning techniques have their limitations in terms of the presence of noticeable artifacts in the injected samples or their vulnerability to preprocessing transformations. Similarly, most defence techniques have their limitations in terms of robustness to different types of poisoning attacks. To overcome these challenges, this research presents a structurally embedded invisible poisoning attack technique and a Holistic Defence Strategy (HDS) for the Pyramid Vision Transformer (PVT) model. In this research, the proposed invisible poisoning attack technique takes advantage of the structural characteristics of images, specifically the edges of images, as invisible carriers of trigger information. The proposed attack technique develops a Deep Multi-Scale U-Net Injection Network (DMS-UNet-IN) to embed the trigger information in images in an invisible manner. The proposed attack technique differs significantly from traditional trigger-based techniques in terms of the alignment of perturbations with structural manifolds. The Holistic Defence Strategy (HDS) develops a discriminative detection boundary in the feature space using a mimic model with an attention-aware generative adversarial network. The proposed invisible poisoning attack technique was evaluated using Radiology, Ophthalmology, and Pathology datasets. The experimental evaluation of the proposed invisible poisoning attack technique showed its superiority in terms of invisibility, with a PSNR value of 43.46 dB and SSIM value of 0.9925, over the state-of-the-art techniques such as DeepPoison (40.91 dB) and SPM (38.16 dB). The proposed Holistic Defence Strategy was evaluated using static poisoning rates ranging from 5 to 30%. The experimental evaluation of the proposed Holistic Defence Strategy showed its superiority over the state-of-the-art techniques such as CD, DUTI, and TRIM in terms of detection accuracy up to 95.8% with an F1-score of more than 0.95. The superiority of the proposed technique was also evident in terms of up to 17% improvement over the baseline techniques at higher levels of contamination.

The major goal of this work is to propose a system that can efficiently read robot diners’ facial expressions for feedback. The suggested research built its sys The major goal of this work is to propose a system that can efficiently read robot diners’ facial expressions for feedback. The suggested research built its sys

The research project aims to assess and contrast the predictive accuracy of two distinct machine learning algorithms: the Decision Tree Classifier (DTC) and a C The research project aims to assess and contrast the predictive accuracy of two distinct machine learning algorithms: the Decision Tree Classifier (DTC) and a C

The study’s goal is to create an efficient feedback system for automated and unmanned restaurants that makes use of customers’ facial expressions by training th The study’s goal is to create an efficient feedback system for automated and unmanned restaurants that makes use of customers’ facial expressions by training th

This research evaluates the performance of linear regression (LR) and support vector machines (SVM) with a custom kernel Trick for predicting the availability a This research evaluates the performance of linear regression (LR) and support vector machines (SVM) with a custom kernel Trick for predicting the availability a

The capacity of Improved Support Vector Machine and Multilayer Perceptron to automatically identify song snippets into various musical genres is compared in this study. To balance the classes, under sampling and oversampling are used to build two separate datasets from an existing internet music collection. We compare the performance of two types of models in this study, including an Improved support vector machine (SVM) that is trained end to end to predict the genre label of an audio signal simply based on its spectrogram. The Multilayer Perceptron is the second model (MLP). This study has two groups, each with a sample size of 20 people and a G power of 80%. The studies were carrie d out on the audio set, and the AUC value for an assembler classifier that combined the two techniques was 0.894. The results show that the support vector machine algorithm has a 69.24 percent accuracy and exceeds the multilayer perceptron, which has a 66.24 percent accuracy. An independent sample T test yielded a statistical significance of P = 0.001 (P 0.05) with a 2 - tailed significance of P = 0.001. The results showed that the upgraded support vector machine with a novel feature extraction strategy for music genre categorization outperformed the multilayer perceptron approach in terms of accuracy.

The creation of sophisticated project-level labor demand forecasting models is discussed in this work. For the entire project’s labor force and ten key crafts, a total of 11 forecasting models for manpower need were created. 54 construction projects were used as the sample for the data collection. Multiple linear regression techniques were used to analyze these data in order to build the estimation models. We compare the results of two types of models in this study. First one is Linear Regression (LR) compared with Novel Time Series Analysis (TSA). The findings show that the labor demand for a project is influenced by a number of factors that are related to its characteristics, such as the project’s kind, construction cost, and traits that relate to its complexity. Four out-of-sample projects and numerous diagnostic tests were used to evaluate and validate the regression models that were produced. This paper consists of two groups with sample size 52. Four out-of-sample projects and numerous diagnostic tests were used to evaluate and validate the regression models that were produced. The models have proven to be solid and trustworthy. The experiments are conducted in the laboratory and we report an average value of 3.85, for an ensemble classifier which combines two approaches. From the results it has been observed that Novel Time Series Analysis has more Accuracy than Linear Regression. Demand Univariate time series, multiple regression, and econometric modeling are examples of manpower forecasting methods that have been created and empirically proven in the construction industry. The use of multiple regression and econometric methods to forecast manpower demand depends on a variety of crucial factors, including employment and pay rates.

This study compares support vector machines (SVM) with a novel kernel trick to linear regression to estimate worker supply and demand for timely completion of public sector industry construction. Materials and methods: Compare the results of the two groups of models in this study. Linear regression comes first, followed by support vector machines. The developed regression models were validated using four out-of-sample projects, several diagnostic tests, and tests for timely completion of public sector industry building. In this study, there are two groups, each having a sample size of 52. Results: The experiments are conducted in the laboratory and we report an average value of 2.97, for an ensemble classifier which combines two approaches. From the results it has been observed that Support vector machines (90.06%) have more Accuracy than Linear Regression (64.02%) and obtained a significance of P = 0.001 (P < 0.05) two-tailed based independent sample T test conducted on both the groups. The suggested SVM model offers a method for predicting the labor demand for construction in the coming years. The quality of the daily records that were obtained from the workers will determine how accurate the model is.