Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
What I Learned Reading Three Knowledge Distillation Papers
Published:
A personal reading log covering Hinton (2015), Gou et al. (2021), and DETRDistill (2023) — from the basics of soft targets to query alignment in transformer-based detectors.
The Core of Data Science: Decoding Data’s Structure with Eigenvalues, SVD, and PCA
Published:
A technical deep dive into foundational Linear Algebra tools—Eigendecomposition and SVD—and their synthesis in PCA for dimensionality reduction.
portfolio
detrflow – End-to-End RT-DETR Object Detection Pipeline
A production-ready object detection pipeline built on RT-DETR, achieving AP=47.9 on COCO val2017. Features a deployed HuggingFace Space, FastAPI inference API, and an accompanying arXiv paper on knowledge distillation.
Student Capacity Moderates Knowledge Distillation Effectiveness
Systematic study of knowledge distillation across ResNet teacher-student pairs on CIFAR-10. Identifies student capacity — not teacher-student accuracy gap — as the key moderating factor in KD effectiveness. Accompanied by a peer-reviewed arXiv publication.
Real-Time Object Detection with YOLOv8
Real-time object detection pipeline using YOLOv8n, with bounding box rendering, class labels, and confidence scores on live video streams.
publications
Student Capacity Moderates Knowledge Distillation Effectiveness: A Systematic Study Across ResNet Teacher-Student Pairs on CIFAR-10
Published in arXiv, 2026
Systematic ablation study comparing Logit-KD and Feature-KD across ResNet teacher-student pairs on CIFAR-10. Key findings: student capacity — not teacher-student accuracy gap — is the primary moderator of KD effectiveness; and implementation correctness critically affects Feature-KD performance.
Recommended citation: Yaşar, U. O. (2026). Student Capacity Moderates Knowledge Distillation Effectiveness. arXiv:2605.31191.
Download Paper