zwt233/NC-ALG

NC-ALG: Graph-based Active Learning under Noisy Crowd.

This repository is the official implementation of NC-ALG.

Requirements

To install requirements:

pip install -r requirements.txt

Training

To train the model(s) in the paper:

cd the “example” data

run the python file NC-ALG.py

Results

1. Accuracy comparison:

2. Accuracy under Different Labeling Budgets:

3. Accuracy under Different Noisy Rates:

4. Generalization Evaluation:

5. Ablation Study:

6. Interpretability: