palm-ml/valen

Instance-Dependent Partial Label Learning（NIPS'21）；Variational Label Enhancement for Instance-Dependent Partial Label Learning（TPAMI）

Paper

VALEN is the implementation of our NIPS'21 paper, "Instance-Dependent Partial Label Learning".

MILEN is the implementation of the extension of the NIPS'21 paper, "Variational Label Enhancement for Instance-Dependent Partial Label Learning", which has been accepted by TPAMI.

Installation

---

pip install -r requirements.txt

Run MILEN

---

cd milen

benchmark-instance

kmnist

python -u main.py -gpu 0 -ds kmnist

fmnist

python -u main.py -gpu 0 -ds fmnist

cifar10

python -u main.py -gpu 0 -ds cifar10

cifar100

python -u main.py -gpu 0 -ds cifar100

Run VALEN-journal

---

cd valen-journal

mnist，kmnist，fmnist

bash run_baseline_mnist.sh

cifar10, cifar100

bash run_baseline_cifar.sh

realworld

bash run_baseline_realworld.sh

benchmark-instance

kmnist

python -u main.py -gpu 0 -ds kmnist

fmnist

python -u main.py -gpu 0 -ds fmnist

cifar10

python -u main.py -gpu 0 -ds cifar10

cifar100

python -u main.py -gpu 0 -ds cifar100

Run VALEN-conference

---

cd valen-conference

benchmark-random

mnist

python -u main.py --gpu 0 --bs 256 --partial_type random --dt benchmark --ds mnist --gamma 10 --beta 0.1

kmnist

python -u main.py --gpu 0 --bs 256 --partial_type random --dt benchmark --ds kmnist --gamma 10 --beta 0.1

fmnist

python -u main.py --gpu 0 --bs 256 --partial_type random --dt benchmark --ds fmnist --gamma 10 --beta 0.1

cifar10

python -u main.py --gpu 0 --bs 256 --partial_type random --dt benchmark --ds cifar10 --lr 5e-2 --wd 1e-3 --gamma 10 --beta 0.1

benchmark-instance

mnist

python -u main.py --gpu 0 --bs 256 --partial_type feature --dt benchmark --ds mnist --warm_up 10 --gamma 5 --beta 0.1

kmnist

python -u main.py --gpu 0 --bs 256 --partial_type feature --dt benchmark --ds kmnist --warm_up 10 --gamma 5 --beta 0.1

fmnist

python -u main.py --gpu 0 --bs 256 --partial_type feature --dt benchmark --ds fmnist --warm_up 10 --gamma 5 --beta 0.1

cifar10

python -u main.py --gpu 0 --bs 256 --partial_type feature --dt benchmark --ds cifar10 --lr 5e-2 --wd 1e-3 --warm_up 10 --gamma 10 --beta 0.1 --correct 0.2

realword

lost

python -u main.py --gpu 0 --bs 100 --dt realworld --ds lost --gamma 20 --beta 0.01

MSRCv2

python -u main.py --gpu 0 --bs 100 --dt realworld --ds MSRCv2 --gamma 20 --beta 0.01

BirdSong

python -u main.py --gpu 0 --bs 100 --dt realworld --ds birdac --gamma 20 --beta 0.01

Soccer Player

python -u main.py --gpu 0 --dt realworld --ds spd --gamma 20 --beta 0.01 --correct 0.2

LYN

python -u main.py --gpu 0 --dt realworld --ds LYN --gamma 20 --beta 0.01 --correct 0.2

Data

---

The data is available at https://drive.google.com/drive/folders/1J_68EqOrLN6tA56RcyTgcr1komJB31Y1?usp=sharing