Research & Publications
My published research work and a curated collection of influential papers in deep learning and computer vision.
My Publications
Google ScholarRenewable energy management in smart home environment via forecast embedded scheduling based on Recurrent Trend Predictive Neural Network
Novel neural network approach for renewable energy forecasting and scheduling in smart home environments.
Engagement detection with multi-task training in e-learning environments
Multi-task deep learning system for analyzing video data and assessing subject engagement during educational content.
Publications I Find Interesting
Transformer-Squared: Self-adaptive LLMs
Novel framework for LLMs that can dynamically adapt to different tasks in real-time by selectively adjusting weight matrix components using reinforcement learning.
Internet Explorer: Targeted Representation Learning on the Open Web
Dynamic approach that actively searches the internet and downloads relevant images to train small-scale ML models, moving beyond static datasets.
DINOv3
Self-supervised learning approach for computer vision with Gram anchoring to address feature degradation, creating versatile visual foundation models.
Perception Encoder: The best visual embeddings are not at the output of the network
Novel vision encoder that produces strong embeddings from intermediate network layers through contrastive vision-language training, achieving SOTA results.
RADIOv2.5: Improved Baselines for Agglomerative Vision Foundation Models
Advanced techniques for training vision foundation models using multi-teacher distillation with solutions for resolution shifts and teacher imbalance.