arXiv subjects

Marimuthu Palaniswami

Publications and source records attributed to Marimuthu Palaniswami.

At least 19 recordsLinked to original sources

arXiv2026

NeuroSleepNet: An Explainable Multi-Head Attention-Based Framework With Spatial and Multi-Scale Independent Temporal Context Learning for Automatic Sleep Stage Scoring

Automatic sleep stage scoring is crucial for monitoring sleep quality and sleep-related disorders. However, most existing electroencephalogram (EEG) based deep learning frameworks rely on current and previous input sequences (varying from 4-100 epochs), which increases computational complexity and creates dependency on transition rules. In this study, we propose NeuroSleepNet, a deep-learning framework that classifies sleep stages using only the current input sequence. NeuroSleepNet employs a two-stage representation learning that includes spatial and multi-scale independent temporal context learning. Multi-head self-attention transformer encoders were employed to transform extracted representations into contextual embeddings to improve the distinctiveness of learned features. Additionally, to handle the class imbalances in different sleep stages, a logarithmic scale-based weighting technique was incorporated into the loss function of NeuroSleepNet. Furthermore, an explainability module was integrated into NeuroSleepNet to visualize and interpret class-specific decision patterns. NeuroSleepNet demonstrated state-of-the-art performance, achieving an accuracy of 86.1 percent, a macro-F1 score of 80.8 percent, and a Cohen's kappa of 0.805 on the Sleep-EDF expanded dataset. It achieved 82.0 percent, 76.3 percent, and 0.753 on MESA, 80.5 percent, 76.8 percent, and 0.738 on Physio2018, and 86.7 percent, 80.9 percent, and 0.804 on the SHHS database, respectively, by utilizing only the current sequence. As a result, there is no need for previous long sequences or sequential dependencies, making it computationally efficient and well-suited for long-term automatic sleep monitoring.

eess.SP↗

arXiv2026

Task-Aligned Self-Supervised Learning for Medical Image Analysis: A Task-Oriented Review with Practical Design Guidelines

Self-supervised learning (SSL) is increasingly used in medical image analysis to reduce dependence on costly expert annotations by learning transferable representations from unlabeled data. However, SSL performance depends not only on model architecture but also on whether the self-supervised objective preserves the information required by the downstream clinical task. This review presents a task-oriented synthesis of SSL methods for medical imaging, focusing on how the design of the self-supervised objective interacts with imaging modality, label availability, and downstream performance. We analyze $78$ studies published from 2017 to 2025 and organize them into four paradigms: contrastive, non-contrastive and predictive, generative and reconstruction-based, and hybrid learning. Rather than cataloging methods chronologically, we examine how these paradigms support classification, segmentation, detection, reconstruction, and regression. The evidence suggests that effectiveness is governed by the match among objective, modality, and downstream task rather than by any single strategy. Contrastive objectives favor global discriminative representations suited to classification but may underrepresent localized pathology, whereas spatial-prediction, masked-modeling, and reconstruction objectives better preserve anatomical structure for segmentation and dense prediction. Critically, misaligned objectives can cause negative transfer through shortcut learning on acquisition signatures or augmentation that erases diagnostic signal rather than merely weaker gains. SSL is most beneficial in low-label regimes, but its effectiveness depends on modality-aware augmentation, pathology-preserving corruption, and clinically meaningful evaluation. We conclude with practical design guidelines and open challenges for clinically aligned SSL.

cs.CV↗

arXiv2026

Generative Diffusion Prior Distillation for Long-Context Knowledge Transfer

While traditional time-series classifiers assume full sequences at inference, practical constraints (latency and cost) often limit inputs to partial prefixes. The absence of class-discriminative patterns in partial data can significantly hinder a classifier's ability to generalize. This work uses knowledge distillation (KD) to equip partial time series classifiers with the generalization ability of their full-sequence counterparts. In KD, high-capacity teacher transfers supervision to aid student learning on the target task. Matching with teacher features has shown promise in closing the generalization gap due to limited parameter capacity. However, when the generalization gap arises from training-data differences (full versus partial), the teacher's full-context features can be an overwhelming target signal for the student's short-context features. To provide progressive, diverse, and collective teacher supervision, we propose Generative Diffusion Prior Distillation (GDPD), a novel KD framework that treats short-context student features as degraded observations of the target full-context features. Inspired by the iterative restoration capability of diffusion models, we learn a diffusion-based generative prior over teacher features. Leveraging this prior, we posterior-sample target teacher representations that could best explain the missing long-range information in the student features and optimize the student features to be minimally degraded relative to these targets. GDPD provides each student feature with a distribution of task-relevant long-context knowledge, which benefits learning on the partial classification task. Extensive experiments across earliness settings, datasets, and architectures demonstrate GDPD's effectiveness for full-to-partial distillation.

cs.LG↗

arXiv2026

MemKD: Memory-Discrepancy Knowledge Distillation for Efficient Time Series Classification

Deep learning models, particularly recurrent neural networks and their variants, such as long short-term memory, have significantly advanced time series data analysis. These models capture complex, sequential patterns in time series, enabling real-time assessments. However, their high computational complexity and large model sizes pose challenges for deployment in resource-constrained environments, such as wearable devices and edge computing platforms. Knowledge Distillation (KD) offers a solution by transferring knowledge from a large, complex model (teacher) to a smaller, more efficient model (student), thereby retaining high performance while reducing computational demands. Current KD methods, originally designed for computer vision tasks, neglect the unique temporal dependencies and memory retention characteristics of time series models. To this end, we propose a novel KD framework termed Memory-Discrepancy Knowledge Distillation (MemKD). MemKD leverages a specialized loss function to capture memory retention discrepancies between the teacher and student models across subsequences within time series data, ensuring that the student model effectively mimics the teacher model's behaviour. This approach facilitates the development of compact, high-performing recurrent neural networks suitable for real-time, time series analysis tasks. Our extensive experiments demonstrate that MemKD significantly outperforms state-of-the-art KD methods. It reduces parameter size and memory usage by approximately 500 times while maintaining comparable performance to the teacher model.

cs.LG↗

arXiv2026

Learning to Reason: Temporal Saliency Distillation for Interpretable Knowledge Transfer

Knowledge distillation has proven effective for model compression by transferring knowledge from a larger network called the teacher to a smaller network called the student. Current knowledge distillation in time series is predominantly based on logit and feature aligning techniques originally developed for computer vision tasks. These methods do not explicitly account for temporal data and fall short in two key aspects. First, the mechanisms by which the transferred knowledge helps the student model learning process remain unclear due to uninterpretability of logits and features. Second, these methods transfer only limited knowledge, primarily replicating the teacher predictive accuracy. As a result, student models often produce predictive distributions that differ significantly from those of their teachers, hindering their safe substitution for teacher models. In this work, we propose transferring interpretable knowledge by extending conventional logit transfer to convey not just the right prediction but also the right reasoning of the teacher. Specifically, we induce other useful knowledge from the teacher logits termed temporal saliency which captures the importance of each input timestep to the teacher prediction. By training the student with Temporal Saliency Distillation we encourage it to make predictions based on the same input features as the teacher. Temporal Saliency Distillation requires no additional parameters or architecture specific assumptions. We demonstrate that Temporal Saliency Distillation effectively improves the performance of baseline methods while also achieving desirable properties beyond predictive accuracy. We hope our work establishes a new paradigm for interpretable knowledge distillation in time series analysis.

cs.LG↗

arXiv2023

MOMA:Distill from Self-Supervised Teachers

Contrastive Learning and Masked Image Modelling have demonstrated exceptional performance on self-supervised representation learning, where Momentum Contrast (i.e., MoCo) and Masked AutoEncoder (i.e., MAE) are the state-of-the-art, respectively. In this work, we propose MOMA to distill from pre-trained MoCo and MAE in a self-supervised manner to collaborate the knowledge from both paradigms. We introduce three different mechanisms of knowledge transfer in the propsoed MOMA framework. : (1) Distill pre-trained MoCo to MAE. (2) Distill pre-trained MAE to MoCo (3) Distill pre-trained MoCo and MAE to a random initialized student. During the distillation, the teacher and the student are fed with original inputs and masked inputs, respectively. The learning is enabled by aligning the normalized representations from the teacher and the projected representations from the student. This simple design leads to efficient computation with extremely high mask ratio and dramatically reduced training epochs, and does not require extra considerations on the distillation target. The experiments show MOMA delivers compact student models with comparable performance to existing state-of-the-art methods, combining the power of both self-supervised learning paradigms. It presents competitive results against different benchmarks in computer vision. We hope our method provides an insight on transferring and adapting the knowledge from large-scale pre-trained models in a computationally efficient way.

cs.CV↗

arXiv2022

Masked Contrastive Representation Learning

Masked image modelling (e.g., Masked AutoEncoder) and contrastive learning (e.g., Momentum Contrast) have shown impressive performance on unsupervised visual representation learning. This work presents Masked Contrastive Representation Learning (MACRL) for self-supervised visual pre-training. In particular, MACRL leverages the effectiveness of both masked image modelling and contrastive learning. We adopt an asymmetric setting for the siamese network (i.e., encoder-decoder structure in both branches), where one branch with higher mask ratio and stronger data augmentation, while the other adopts weaker data corruptions. We optimize a contrastive learning objective based on the learned features from the encoder in both branches. Furthermore, we minimize the $L_1$ reconstruction loss according to the decoders' outputs. In our experiments, MACRL presents superior results on various vision benchmarks, including CIFAR-10, CIFAR-100, Tiny-ImageNet, and two other ImageNet subsets. Our framework provides unified insights on self-supervised visual pre-training and future research.

cs.CV↗

arXiv2022

Scheduling IoT Applications in Edge and Fog Computing Environments: A Taxonomy and Future Directions

Fog computing, as a distributed paradigm, offers cloud-like services at the edge of the network with low latency and high-access bandwidth to support a diverse range of IoT application scenarios. To fully utilize the potential of this computing paradigm, scalable, adaptive, and accurate scheduling mechanisms and algorithms are required to efficiently capture the dynamics and requirements of users, IoT applications, environmental properties, and optimization targets. This paper presents a taxonomy of recent literature on scheduling IoT applications in Fog computing. Based on our new classification schemes, current works in the literature are analyzed, research gaps of each category are identified, and respective future directions are described.

cs.DC↗

arXiv2022

Achieving AI-enabled Robust End-to-End Quality of Experience over Radio Access Networks

Emerging applications such as Augmented Reality, the Internet of Vehicles and Remote Surgery require both computing and networking functions working in harmony. The End-to-end (E2E) quality of experience (QoE) for these applications depends on the synchronous allocation of networking and computing resources. However, the relationship between the resources and the E2E QoE outcomes is typically stochastic and non-linear. In order to make efficient resource allocation decisions, it is essential to model these relationships. This article presents a novel machine-learning based approach to learn these relationships and concurrently orchestrate both resources for this purpose. The machine learning models further help make robust allocation decisions regarding stochastic variations and simplify robust optimization to a conventional constrained optimization. When resources are insufficient to accommodate all application requirements, our framework supports executing some of the applications with minimal degradation (graceful degradation) of E2E QoE. We also show how we can implement the learning and optimization methods in a distributed fashion by the Software-Defined Network (SDN) and Kubernetes technologies. Our results show that deep learning-based modelling achieves E2E QoE with approximately 99.8\% accuracy, and our robust joint-optimization technique allocates resources efficiently when compared to existing differential services alternatives.

cs.NI↗

arXiv2022

Online Slice Reconfiguration for End-to-End QoE in 6G Applications

End-to-end (E2E) quality of experience (QoE) for 6G applications depends on the synchronous allocation of networking and computing resources, also known as slicing. However, the relationship between the resources and the E2E QoE outcomes is typically stochastic and non-stationary. Existing works consider known resource demands for slicing and formulate optimization problems for slice reconfiguration. In this work, we create and manage slices by learning the relationship between E2E QoE and resources. We develop a gradient-based online slice reconfiguration algorithm (OSRA) to reconfigure and manage slices in resource-constrained scenarios for radio access networks (RAN). We observe that our methodology meets the QoE requirements with high accuracy compared to existing approaches. It improves upon the existing approaches by approximately 98\% for bursty traffic variations. Our algorithm has fast convergence and achieves low E2E delay violations for lower priority slices.

cs.NI↗

arXiv2021

A Distributed Deep Reinforcement Learning Technique for Application Placement in Edge and Fog Computing Environments

Fog/Edge computing is a novel computing paradigm supporting resource-constrained Internet of Things (IoT) devices by the placement of their tasks on the edge and/or cloud servers. Recently, several Deep Reinforcement Learning (DRL)-based placement techniques have been proposed in fog/edge computing environments, which are only suitable for centralized setups. The training of well-performed DRL agents requires manifold training data while obtaining training data is costly. Hence, these centralized DRL-based techniques lack generalizability and quick adaptability, thus failing to efficiently tackle application placement problems. Moreover, many IoT applications are modeled as Directed Acyclic Graphs (DAGs) with diverse topologies. Satisfying dependencies of DAG-based IoT applications incur additional constraints and increase the complexity of placement problems. To overcome these challenges, we propose an actor-critic-based distributed application placement technique, working based on the IMPortance weighted Actor-Learner Architectures (IMPALA). IMPALA is known for efficient distributed experience trajectory generation that significantly reduces the exploration costs of agents. Besides, it uses an adaptive off-policy correction method for faster convergence to optimal solutions. Our technique uses recurrent layers to capture temporal behaviors of input data and a replay buffer to improve the sample efficiency. The performance results, obtained from simulation and testbed experiments, demonstrate that our technique significantly improves the execution cost of IoT applications up to 30\% compared to its counterparts.

cs.DC↗

arXiv2021

Achieving QoS for Real-Time Bursty Applications over Passive Optical Networks

Emerging real-time applications such as those classified under ultra-reliable low latency (uRLLC) generate bursty traffic and have strict Quality of Service (QoS) requirements. Passive Optical Network (PON) is a popular access network technology, which is envisioned to handle such applications at the access segment of the network. However, the existing standards cannot handle strict QoS constraints. The available solutions rely on instantaneous heuristic decisions and maintain QoS constraints (mostly bandwidth) in an average sense. Existing works with optimal strategies are computationally complex and are not suitable for uRLLC applications. This paper presents a novel computationally-efficient, far-sighted bandwidth allocation policy design for facilitating bursty traffic in a PON framework while satisfying strict QoS (age of information/delay and bandwidth) requirements of modern applications. To this purpose, first we design a delay-tracking mechanism which allows us to model the resource allocation problem from a control-theoretic viewpoint as a Model Predictive Control (MPC). MPC helps in taking far-sighted decisions regarding resource allocations and captures the time-varying dynamics of the network. We provide computationally efficient polynomial-time solutions and show its implementation in the PON framework. Compared to existing approaches, MPC reduces delay violations by approximately 15% for a delay-constrained application of 1ms target. Our approach is also robust to varying traffic arrivals.

cs.NI↗

arXiv2021

A Distributed Application Placement and Migration Management Techniques for Edge and Fog Computing Environments

Fog/Edge computing model allows harnessing of resources in the proximity of the Internet of Things (IoT) devices to support various types of real-time IoT applications. However, due to the mobility of users and a wide range of IoT applications with different requirements, it is a challenging issue to satisfy these applications' requirements. The execution of IoT applications exclusively on one fog/edge server may not be always feasible due to limited resources, while execution of IoT applications on different servers needs further collaboration among servers. Also, considering user mobility, some modules of each IoT application may require migration to other servers for execution, leading to service interruption and extra execution costs. In this article, we propose a new weighted cost model for hierarchical fog computing environments, in terms of the response time of IoT applications and energy consumption of IoT devices, to minimize the cost of running IoT applications and potential migrations. Besides, a distributed clustering technique is proposed to enable the collaborative execution of tasks, emitted from application modules, among servers. Also, we propose an application placement technique to minimize the overall cost of executing IoT applications on multiple servers in a distributed manner. Furthermore, a distributed migration management technique is proposed for the potential migration of applications' modules to other remote servers as the users move along their path. Besides, failure recovery methods are embedded in the clustering, application placement, and migration management techniques to recover from unpredicted failures. The performance results show that our technique significantly improves its counterparts in terms of placement deployment time, average execution cost of tasks, total number of migrations, total number of interrupted tasks, and cumulative migration cost.

cs.DC↗

arXiv2019

A Scalable Framework for Trajectory Prediction

Trajectory prediction (TP) is of great importance for a wide range of location-based applications in intelligent transport systems such as location-based advertising, route planning, traffic management, and early warning systems. In the last few years, the widespread use of GPS navigation systems and wireless communication technology enabled vehicles has resulted in huge volumes of trajectory data. The task of utilizing this data employing spatio-temporal techniques for trajectory prediction in an efficient and accurate manner is an ongoing research problem. Existing TP approaches are limited to short-term predictions. Moreover, they cannot handle a large volume of trajectory data for long-term prediction. To address these limitations, we propose a scalable clustering and Markov chain based hybrid framework, called Traj-clusiVAT-based TP, for both short-term and long-term trajectory prediction, which can handle a large number of overlapping trajectories in a dense road network. Traj-clusiVAT can also determine the number of clusters, which represent different movement behaviours in input trajectory data. In our experiments, we compare our proposed approach with a mixed Markov model (MMM)-based scheme, and a trajectory clustering, NETSCAN-based TP method for both short- and long-term trajectory predictions. We performed our experiments on two real, vehicle trajectory datasets, including a large-scale trajectory dataset consisting of 3.28 million trajectories obtained from 15,061 taxis in Singapore over a period of one month. Experimental results on two real trajectory datasets show that our proposed approach outperforms the existing approaches in terms of both short- and long-term prediction performances, based on prediction accuracy and distance error (in km).

cs.AI↗

arXiv2018

The effects of asymmetric volume conductor modeling on non-invasive fetal ECG extraction

Objective: Non-invasive fetal electrocardiography (NI-FECG) shows promise for capturing novel physiological information that may indicate signs of fetal distress. However, significant deterioration in NI-FECG signal quality occurs during the presence of a highly non-conductive layer known as vernix caseosa which forms on the fetal body surface beginning in approximately the 28th week of gestation. This work investigates asymmetric modeling of vernix caseosa and other maternal-fetal tissues in accordance with clinical observations and assesses their impacts for NI-FECG signal processing. Methods: We develop a process for simulating dynamic maternal-fetal abdominal ECG mixtures using a synthetic cardiac source model embedded in a finite element volume conductor. Using this process, changes in NI-FECG signal morphology are assessed in an extensive set of finite element models including spatially variable distributions of vernix caseosa. Results: Our simulations show that volume conductor asymmetry can result in over 70% error in the observed T/QRS ratio and significant changes to signal morphology compared to a homogeneous volume conductor model. Conclusion: Volume conductor effects must be considered when analyzing T/QRS ratios obtained via NI-FECG and should be considered in future algorithm benchmarks using simulated data. Significance: This work shows that without knowledge of the influence of volume conductor effects, clinical evaluation of the T/QRS ratio derived via NI-FECG should be avoided.

physics.med-ph↗

arXiv2016

Fuzzy c-Shape: A new algorithm for clustering finite time series waveforms

The existence of large volumes of time series data in many applications has motivated data miners to investigate specialized methods for mining time series data. Clustering is a popular data mining method due to its powerful exploratory nature and its usefulness as a preprocessing step for other data mining techniques. This article develops two novel clustering algorithms for time series data that are extensions of a crisp c-shapes algorithm. The two new algorithms are heuristic derivatives of fuzzy c-means (FCM). Fuzzy c-Shapes plus (FCS+) replaces the inner product norm in the FCM model with a shape-based distance function. Fuzzy c-Shapes double plus (FCS++) uses the shape-based distance, and also replaces the FCM cluster centers with shape-extracted prototypes. Numerical experiments on 48 real time series data sets show that the two new algorithms outperform state-of-the-art shape-based clustering algorithms in terms of accuracy and efficiency. Four external cluster validity indices (the Rand index, Adjusted Rand Index, Variation of Information, and Normalized Mutual Information) are used to match candidate partitions generated by each of the studied algorithms. All four indices agree that for these finite waveform data sets, FCS++ gives a small improvement over FCS+, and in turn, FCS+ is better than the original crisp c-shapes method. Finally, we apply two tests of statistical significance to the three algorithms. The Wilcoxon and Friedman statistics both rank the three algorithms in exactly the same way as the four cluster validity indices.

cs.LG↗

arXiv2012

Internet of Things (IoT): A Vision, Architectural Elements, and Future Directions

Ubiquitous sensing enabled by Wireless Sensor Network (WSN) technologies cuts across many areas of modern day living. This offers the ability to measure, infer and understand environmental indicators, from delicate ecologies and natural resources to urban environments. The proliferation of these devices in a communicating-actuating network creates the Internet of Things (IoT), wherein, sensors and actuators blend seamlessly with the environment around us, and the information is shared across platforms in order to develop a common operating picture (COP). Fuelled by the recent adaptation of a variety of enabling device technologies such as RFID tags and readers, near field communication (NFC) devices and embedded sensor and actuator nodes, the IoT has stepped out of its infancy and is the the next revolutionary technology in transforming the Internet into a fully integrated Future Internet. As we move from www (static pages web) to web2 (social networking web) to web3 (ubiquitous computing web), the need for data-on-demand using sophisticated intuitive queries increases significantly. This paper presents a cloud centric vision for worldwide implementation of Internet of Things. The key enabling technologies and application domains that are likely to drive IoT research in the near future are discussed. A cloud implementation using Aneka, which is based on interaction of private and public clouds is presented. We conclude our IoT vision by expanding on the need for convergence of WSN, the Internet and distributed computing directed at technological research community.

cs.DC↗

arXiv2012

Cramér-Rao Bounds for Polynomial Signal Estimation using Sensors with AR(1) Drift

We seek to characterize the estimation performance of a sensor network where the individual sensors exhibit the phenomenon of drift, i.e., a gradual change of the bias. Though estimation in the presence of random errors has been extensively studied in the literature, the loss of estimation performance due to systematic errors like drift have rarely been looked into. In this paper, we derive closed-form Fisher Information matrix and subsequently Cramér-Rao bounds (upto reasonable approximation) for the estimation accuracy of drift-corrupted signals. We assume a polynomial time-series as the representative signal and an autoregressive process model for the drift. When the Markov parameter for drift ρ<1, we show that the first-order effect of drift is asymptotically equivalent to scaling the measurement noise by an appropriate factor. For ρ=1, i.e., when the drift is non-stationary, we show that the constant part of a signal can only be estimated inconsistently (non-zero asymptotic variance). Practical usage of the results are demonstrated through the analysis of 1) networks with multiple sensors and 2) bandwidth limited networks communicating only quantized observations.

cs.IT↗