Incremental Learning
Incremental learning aims to develop artificial intelligence systems that can continuously learn from new data to address new tasks while retaining knowledge acquired from previous tasks. This approach updates the model constantly, enabling it to adapt to new environments without forgetting old knowledge, thereby enhancing the system's long-term adaptability and efficiency, which holds significant application value.
CIFAR-100 - 50 classes + 5 steps of 10 classes
PODNet (CNN)
CIFAR-100 - 50 classes + 10 steps of 5 classes
DER(Standard ResNet-18)
ImageNet100 - 10 steps
DER w/o Pruning
ImageNet - 10 steps
CIFAR-100-B0(5steps of 20 classes)
CIFAR-100 - 50 classes + 25 steps of 2 classes
RMM (Modified ResNet-32)
CIFAR100-B0(10steps of 10 classes)
ImageNet-100 - 50 classes + 10 steps of 5 classes
RMM (ResNet-18)
ImageNet-100 - 50 classes + 5 steps of 10 classes
RMM (ResNet-18)
CIFAR-100 - 50 classes + 2 steps of 25 classes
TCIL
CIFAR100B020Step(5ClassesPerStep)
ImageNet - 500 classes + 5 steps of 100 classes
RMM (ResNet-18)
ImageNet - 500 classes + 10 steps of 50 classes
PODNet
ImageNet-100 - 50 classes + 25 steps of 2 classes
CIFAR-100 - 50 classes + 50 steps of 1 class
PODNet
ImageNet-100 - 50 classes + 50 steps of 1 class
PODNet
CIFAR-100 - 40 classes + 60 steps of 1 class (Exemplar-free)
FeTrIL
CIFAR100B050S(2ClassesPerStep)
DER(ResNet-18)
ImageNet-10k - 5225 classes + 5 steps of 1045 classes
PPCA-CLIP
ImageNet - 500 classes + 25 steps of 20 classes
RMM (ResNet-18)
ImageNet100 - 20 steps
FOSTER
MLT17
MRM