12 datasets found

A variant of CIFAR-100 dataset, where the original dataset is split into 20 disjoint tasks, each consisting of 2,500 samples from 5 classes.

Generative kernel continual learning dataset

Continual Test Time Adaptation (CTTA) is re-required to adapt efficiently to continuous unseen domains while retaining previously learned knowledge.

The dataset used in the paper is a continual learning benchmark, consisting of three datasets: splitMNIST, splitFashionMNIST, and splitCIFAR.

The Sparse Diffusion Policy (SDP) framework, which integrates Mixture of Experts (MoE) layers into the diffusion policy.

Real-world time-series benchmark datasets for crude oil prices with distribution shifts

The dataset used in this paper for Continual Learning, Catastrophic Forgetting and PCA-OGD.

The Permuted MNIST dataset is a variation of MNIST where new tasks of comparable difficulty to the original MNIST classification task are created by permuting the pixels of...

The dataset used in this paper for Continual Learning, Catastrophic Forgetting and PCA-OGD.

The Rotated MNIST dataset is a subset of the MNIST dataset with images of handwritten digits rotated by 90 degrees.

Few-Shot Class Incremental Learning (FSCIL) is a challenging continual learning task, where limited training examples are available during several learning sessions.

CORe50 is a video benchmark for continual learning, consisting of 164,866 128x128 images of 50 domestic objects belonging to 10 categories.