Aditya Birla Group

We present the first causal mechanistic analysis of a tabular foundation model, investigating how TabPFN 2.5's feature wise attention heads distribute computation across layers. Using activation patching, ablation, and attention entropy across two synthetic regression datasets, we find clear temporal specialisation: one head's causal necessity dominates that of the others by 2 to 5 times at peak layer, with its dominant layer shifting across tasks of different complexity, while the remaining heads exhibit symmetric late layer profiles. Attention entropy and patching provide convergent evidence for the computationally active layers of the dominant head. We additionally investigate inference time steerability via contrastive activation steering, which fails to transfer across samples. We attribute this result to TabPFN's in context learning mechanism, which encodes task structure through context dependent attention rather than the stable parametric directions that make steering tractable in language models.

The study employs a novel trust-aided routing approach that employs a stacked gated recurrent unit (SGRU) to estimate the node's trustworthiness in mobile ad-hoc networks (MANETs). In this designed model, an optimal path for effective routing is selected by the hybrid golf optimizer with walruses behavior-based optimization algorithm (HGO-WBOA), which is the combination of both golf optimization algorithm (GOA) and walruses behavior optimization algorithm (WBOA). The effectiveness of routing is enhanced by considering some multi-object constraints such as remaining energy, throughput, trust, path loss, packet energy cost, security, battery, routing overhead ratio, bandwidth, link cost, end-to-end delay, hop count, and packet delivery ratio (PDR) during the optimization process. This designed SGRU identifies security threats and provides an alternate node by removing the attack node to guarantee the integrity of the network. The numerical validations are performed with some traditional energy efficiency routing approaches in MANET. The throughput of the recommended HGO-WBOA-SGRU is 95%, which is greater than the previous approaches, such as COA-SGRU (88.5%), LOA-SGRU (94.5%), GOA-SGRU (84%), and WBO-SGRU (93%), respectively, for the 200th node value. The outcomes prove the effectiveness of the designed approach in recognizing and reducing security threats while also enhancing energy efficiency and routing performance. The novelty of the proposed approach lies in its integration of SGRU with trust-aided routing, enabling adaptive and accurate trust estimations in dynamic network conditions. This innovative framework allows the model to capture complex relations among nodes and adapt to dynamic network conditions, offering reliable and robust solutions for secure routing in MANETs.

The present study investigates the optical performance of a V-trough solar concentrator using four distinct bottom reflector configurations: flat, parabolic (PP), circular (CC), and a combination of circular and parabolic (CC–PP). Detailed ray tracing simulations were conducted to analyze the distribution of heat flux, optical efficiency, and local concentration ratio (LCR) across varying incidence angles for each configuration. The analysis compares bifacial and trifacial absorbers using tonatiuh ray tracing software. The novel trifacial absorber features three active surfaces arranged in a triangular layout, which enhances the light capture and optical efficiency. The left and right lateral surfaces of the trifacial absorber exhibit better uniformity and local concentration ratio across all incident angles. The trifacial system outperformed the bifacial system with all types of reflector configurations (flat, CC, PP, and CC–PP). The CC–PP achieved the maximum optical efficiency of 85%, 66%, and 40% (trifacial) and 46%, 53%, and 38% (bifacial) at incident angles of 0 deg, 20 deg, and 40 deg, respectively, followed by PP, CC, and flat configurations. The same trend follows with other incidence angles. Among all bottom reflector configurations, the CC–PP exhibited better performance, reaching a peak value of local concentration ratio of 4.5 at 30 deg and a maximum heat flux of 4500 W/m2 on the bottom surface.

Mobile networks have an ever-increasing presence in critical infrastructure, enabling just-in-time monitoring and real-time control in various domains, from energy and transport to healthcare. However, the security of these systems is now vulnerable to new cybersecurity threats from this integration. Evaluating the potential of behavioral analytics and anomaly detection in securing mobile networks for critical infrastructure. This paper presents a novel Multi-Layer Adaptive Anomaly Detection System (MAADS) that utilizes recent advances in machine learning (ML) to identify anomalies and detect novel threats in mobile networks. MAADS detects anomalies in mobile networks across three layers: network traffic, user mobility, and device behavior. It does this through privacy-preserving data collection, multi-model anomaly detection, and an adaptive response framework. The empirical evaluation of the proposed system shows that MAADS can detect anomalies with 95% precision and 93% recall. It is demonstrated to identify 87% of anomalies not detected by standard ML-based systems. It also maintains its performance up to 100,000 nodes and consistently performs across other critical infrastructure sectors. The analysis and comparison of MAADS with existing solutions show that it outperforms existing techniques in terms of the ability to adapt to novel threats and the explainability of the detected anomalies.

The rapid increase in cardiovascular diseases has necessitated the development of intelligent, scalable, and realtime healthcare solutions capable of early diagnosis and prevention. Smart healthcare systems integrating the Internet of Things (IoT) and Machine Learning (ML) have emerged as transformative technologies that enable continuous monitoring, data-driven decision-making, and predictive analytics in clinical environments. This study presents a comprehensive framework for heart disease prediction using IoT-enabled sensing devices and advanced machine learning algorithms, emphasizing real-world applicability and algorithmic development. The proposed system leverages wearable and embedded sensors to collect physiological parameters such as heart rate, blood pressure, and electrocardiogram signals, which are transmitted through cloud-based architectures for preprocessing and analysis. Machine learning models, including supervised and ensemble approaches, are developed to identify patterns and predict cardiovascular risk with high accuracy. The study further explores optimization strategies, feature selection techniques, and model interpretability to enhance predictive performance and clinical reliability. Real-world implementation scenarios are analyzed to demonstrate the feasibility of integrating such systems into modern healthcare infrastructures, including remote patient monitoring and telemedicine platforms. Additionally, the research highlights the challenges associated with data quality, privacy, interoperability, and scalability while proposing solutions for robust deployment. The findings indicate that IoT and ML-based healthcare systems significantly improve early diagnosis, reduce mortality rates, and support personalized treatment strategies. This research contributes to the advancement of intelligent healthcare by bridging the gap between theoretical models and practical applications in heart disease prediction.