A5012487013- Synthesis and biological activity
- Domain Adaptation and Few-Shot Learning
- Luminescence Properties of Advanced Materials
- Nonlinear Optical Materials Research
- Synthesis and Characterization of Heterocyclic Compounds
- Generative Adversarial Networks and Image Synthesis
- Internet Traffic Analysis and Secure E-voting
- Advanced machining processes and optimization
- Thin-Film Transistor Technologies
- Network Traffic and Congestion Control
- Advanced Machining and Optimization Techniques
- Radiation Detection and Scintillator Technologies
- Advanced Surface Polishing Techniques
- Microgrid Control and Optimization
- Caching and Content Delivery
- Solar Thermal and Photovoltaic Systems
- Multicomponent Synthesis of Heterocycles
- Face and Expression Recognition
- Organic Electronics and Photovoltaics
- Conducting polymers and applications
- Aluminum Alloys Composites Properties
- Metal complexes synthesis and properties
- Network Security and Intrusion Detection
- Multiferroics and related materials
- Chaos-based Image/Signal Encryption
Chandigarh University
2023-2025
Laboratoire de Chimie et Physique Quantiques
2025
Université de Sherbrooke
2025
Geological Survey of India
2024
University of Lucknow
2015-2024
Bihar Agricultural University
2019-2024
Motilal Nehru National Institute of Technology
2016-2024
University of Canterbury
2024
Pacific Northwest National Laboratory
2023-2024
Tata Consultancy Services (India)
2024
Early Separation In photovoltaic devices, electrons excited by the absorption of light must travel across a junction, while positively charged “holes” they leave behind effectively migrate in opposite direction. If and holes do not separate efficiently, can recombine fail to produce any appreciable current. Gélinas et al. (p. 512 , published online 12 December; see Perspective Bredas ) studied this separation process ultrafast optical spectroscopy thiophene-derived donor-fullerene acceptor...
Creating noise from data is easy; creating generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms complex distribution to known prior by slowly injecting noise, and corresponding reverse-time SDE the back into removing noise. Crucially, depends only on time-dependent gradient field (\aka, score) of perturbed distribution. By leveraging advances in score-based modeling, we can accurately estimate these scores with neural networks, use numerical...
Very deep convolutional neural networks offer excellent recognition results, yet their computational expense limits impact for many real-world applications. We introduce BlockDrop, an approach that learns to dynamically choose which layers of a network execute during inference so as best reduce total computation without degrading prediction accuracy. Exploiting the robustness Residual Networks (ResNets) layer dropping, our framework selects on-the-fly residual blocks evaluate given novel...
Transfer learning, which allows a source task to affect the inductive bias of target task, is widely used in computer vision. The typical way conducting transfer learning with deep neural networks fine-tune model pretrained on using data from task. In this paper, we propose an adaptive fine-tuning approach, called SpotTune, finds optimal strategy per instance for data. given image policy network make routing decisions whether pass through fine-tuned layers or pre-trained layers. We conduct...
Multi-task learning aims to improve generalization performance of multiple prediction tasks by appropriately sharing relevant information across them. In the context deep neural networks, this idea is often realized hand-designed network architectures with layers that are shared and branches encode task-specific features. However, space possible multi-task combinatorially large final architecture arrived at manual exploration space, which can be both error-prone tedious. We propose an...
Physical forces in the form of substrate rigidity or geometrical constraints have been shown to alter gene expression profile and differentiation programs. However, underlying mechanism regulation by these mechanical cues is largely unknown. In this work, we use micropatterned substrates cellular geometry (shape, aspect ratio, size) study nuclear mechanotransduction regulate expression. Genome-wide transcriptome analysis revealed cell geometry-dependent alterations actin-related Increase...
Efforts to tune the bulk physical properties of concrete are hindered by a lack knowledge related atomic-level structure and growth calcium silicate hydrate phases, which form about 50–60% volume cement paste. Here we describe first synthesis compositionally uniform phases with Ca:Si ratios tunable between 1.0 2.0. The synthesized here does not contain secondary Ca(OH)2 phase, even in samples above 1.6, is unprecedented for synthetic systems. We then solve three-dimensional these materials...
Learning to classify new categories based on just one or a few examples is long-standing challenge in modern computer vision. In this work, we proposes simple yet effective method for few-shot (and one-shot) object recognition. Our approach modified auto-encoder, denoted Delta-encoder, that learns synthesize samples an unseen category by seeing from it. The synthesized are then used train classifier. proposed both extract transferable intra-class deformations, "deltas", between same-class...
Heart failure is a frequent cause of hospitalization and readmission because the severity disease. Researchers explored using Machine Learning (ML) algorithms to forecast whether heart patients must be readmitted hospital. This study examines ML that use data from electronic health records hospital readmissions for with failure. We will assess accuracy, precision, recall, F1-score logistic regression, decision trees, random forests, Support Vector Machines (SVM), artificial neural networks....
Transfer learning is a popular deep technique that involves fine-tuning pre-trained CNN model on new dataset to improve accuracy and speed. This article examines the effectiveness of transfer techniques in image classification tasks using CNNs. The paper reviews recent studies techniques, including their use medical analysis applications such as COVID-19 detection Alzheimer's disease classification. study discusses ImageNet benchmark for pretraining models proposes an optimized uses various...
This paper explores the multifaceted nature of β-glucan, a notable dietary fiber (DF) with extensive applications. Beginning an in-depth examination its intricate polysaccharide structure, discussion extends to diverse sources like oats, barley, mushrooms, and yeast, emphasizing their unique compositions. The absorption metabolism β-glucan in human body are scrutinized, potential health benefits. Extraction purification processes for high-quality food, pharmaceuticals, cosmetics outlined....
Knowing the distribution of sizes traffic flows passing through a network link helps operator to characterize resource usage, infer demands, detect anomalies, and accommodate new demands better engineering. Previous work on estimating flow size has been focused making inferences from sampled traffic. Its accuracy is limited by (typically) low sampling rate required make operation affordable. In this paper we present novel data streaming algorithm provide much more accurate estimates...
A simple technique was developed to fabricate a large-area TiO2 electrode layer using electrospun nanorods for dye-sensitized solar cells (DSSCs). Using this technique, we assembled DSSCs of area ∼1 cm2 consisting thin nanoparticle and thick nanorod as electrode. The were obtained by mechanically grinding nanofibers. titania sol first spin-coated on conductive glass plate next spray dried it nanoparticle/nanorod layers. These layers subsequently sintered. best-performing DSSC evaluated under...
The anatase TiO2 nanofibers of average diameters 60, 100, and 150 nm were fabricated by controlled electrospinning a polymeric solution subsequent sintering the as-spun fibers. sintered fibers polycrystalline composed densely packed grains size ∼12 nm. rutile phase nucleated at particle interface dense temperature <570 °C because increased surface stress observed in these nanofibers. X-ray electron diffraction measurements analysis showed that lattice strain with decrease fiber diameter....