03 / SELECTED WORKMachine learning / Research
Noisy labels.
Hybrid Deep Learning for Noisy Label Classification
Implemented and compared four label-noise robustness strategies under a fixed ResNet-18 pipeline on CIFAR-10N human-annotated labels.
CIFAR-10N / RESNET-18
87.53%
Peak validation accuracy
86.91%Baseline
04Strategies compared
2.4ppLate-epoch decay eliminated
Learning better from imperfect labels.
01 Overview
This project implements and compares four label-noise robustness strategies — cross-entropy baseline, neighbor-consistency regularization, agreement-based hybrid reweighting, and contrastive pretraining — under a fixed ResNet-18 pipeline on CIFAR-10N human-annotated labels.
02 What I built
- ResNet-18 classification pipeline with four noise-robustness strategies
- Structural and hybrid loss functions combining feature-space neighbor consistency with confidence-based sample reweighting
- Contrastive pretraining workflow to stabilize representation learning
- Diagnostic evaluation pipeline: learning curves, confusion matrices, UMAP/t-SNE embeddings
03 Engineering decisions
- Achieved 87.53% peak validation accuracy on CIFAR-10N vs. 86.91% baseline and eliminated 2.4pp of late-epoch memorization decay
- Designed hybrid loss combining neighbor consistency and confidence-based reweighting to address both noise and representation geometry
- Used UMAP/t-SNE to interpret confusion patterns across visually-ambiguous classes
- Focused on reproducibility with a fixed pipeline to isolate the effect of each robustness strategy