A5001909583- Speech and Audio Processing
- Remote-Sensing Image Classification
- Glaucoma and retinal disorders
- Hearing Loss and Rehabilitation
- Advanced Neural Network Applications
- Sepsis Diagnosis and Treatment
- Machine Learning and ELM
- Advanced Image and Video Retrieval Techniques
- Reinforcement Learning in Robotics
- Retinal Diseases and Treatments
- Animal Vocal Communication and Behavior
- Retinal Imaging and Analysis
- Music and Audio Processing
- Domain Adaptation and Few-Shot Learning
- Clinical Reasoning and Diagnostic Skills
Département d'Informatique
2022
Birla Institute of Technology and Science, Pilani - Dubai Campus
2020-2021
Batch normalization is a staple of computer vision models, including those employed in few-shot learning. nor-malization layers convolutional neural networks are composed step, followed by shift and scale these normalized features applied via the per-channel trainable affine parameters <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\gamma$</tex> xmlns:xlink="http://www.w3.org/1999/xlink">$\beta$</tex> . These param-eters were introduced to...
Transformers have significantly impacted domains like natural language processing, computer vision, and robotics, where they improve performance compared to other neural networks. This survey explores how transformers are used in reinforcement learning (RL), seen as a promising solution for addressing challenges such unstable training, credit assignment, lack of interpretability, partial observability. We begin by providing brief domain overview RL, followed discussion on the classical RL...
Offline reinforcement learning has shown promise for solving tasks in safety-critical settings, such as clinical decision support. Its application, however, been limited by the lack of interpretability and interactivity clinicians. To address these challenges, we propose medical transformer (MeDT), a novel versatile framework based on goal-conditioned paradigm sepsis treatment recommendation. MeDT uses architecture to learn policy drug dosage During offline training, utilizes collected...
Humans have perfected the art of learning from multiple modalities through sensory organs. Despite their impressive predictive performance on a single modality, neural networks cannot reach human level accuracy with respect to modalities. This is particularly challenging task due variations in structure respective Conditional Batch Normalization (CBN) popular method that was proposed learn contextual features aid deep tasks. technique uses auxiliary data improve representational power by...