Association Graph Learning for Multi-Task Classification with Category Shifts

Code for paper “Association Graph Learning for Multi-Task Classification with Category Shifts” accepted to NeurIPS2022.

Set Up

Prerequisites

• Python 3.9.7
• Pytorch 1.11.0
• GPU: an NVIDIA Tesla V100

Getting Started

Inside this repository, we mainly conduct comprehensive experiments on Office-Home. Download the dataset from the following link, and place it in ../../dataset/. To split documents are obtained by randomly selecting 80% of samples from each task as the complete training set and use the remaining samples as the test set. The split documents used for the office-home dataset is provided in train_split/. The class assignment documents with different missing rates (75%, 50%, 25% and 0%) is provided in the tables of the supplemental materials.

• Office-home; [link]

Experiments

To train the proposed association graph by running the command:

python set_up.py --gpu_id 0 --dataset office-home --missing_rate 0.75 --nlayers 4