Command Palette
Search for a command to run...
持续学习
持续学习（Continual Learning），又称增量学习或终身学习，是指在不遗忘先前任务知识的前提下，顺序学习大量任务的模型训练方法。在数据不可复现的情况下，该方法通过提供任务标识来评估验证效果，旨在解决模型在不断变化的任务中保持性能的问题，具有重要的应用价值。
ASC (19 tasks)
CTR
visual domain decathlon (10 tasks)
Res. adapt. decay
Tiny-ImageNet (10tasks)
ALTA-ViTB/16
Cifar100 (20 tasks)
Model Zoo-Continual
F-CelebA (10 tasks)
CAT (CNN backbone)
ImageNet (Fine-grained 6 Tasks)
CondConvContinual
Wikiart (Fine-grained 6 Tasks)
Stanford Cars (Fine-grained 6 Tasks)
CPG
Sketch (Fine-grained 6 Tasks)
20Newsgroup (10 tasks)
Flowers (Fine-grained 6 Tasks)
CondConvContinual
DSC (10 tasks)
CTR
CUBS (Fine-grained 6 Tasks)
CondConvContinual
ImageNet-50 (5 tasks)
CondConvContinual
Cifar100 (10 tasks)
RMN (Resnet)
Permuted MNIST
RMN
split CIFAR-100
Rotated MNIST
Model Zoo-Continual
5-dataset - 1 epoch
5-Datasets
TinyImageNet ResNet-18 - 300 Epochs
CIFAR-100 AlexNet - 300 Epoch
CIFAR-100 ResNet-18 - 300 Epochs
IBM
Split MNIST (5 tasks)
H$^{2}$
Split CIFAR-10 (5 tasks)
H$^{2}$
MLT17
MiniImageNet ResNet-18 - 300 Epochs
miniImagenet
mini-Imagenet (20 tasks) - 1 epoch
TAG-RMSProp
Cifar100 (20 tasks) - 1 epoch
Coarse-CIFAR100
Model Zoo-Continual
CUB-200-2011 (20 tasks) - 1 epoch