Class incremental learning (CIL) has attracted a great deal of attention as an effective way to realize lifelong learning. However, existing works still face catastrophic forgetting, i.e., performance degradation on old tasks after learning new category information. In this work, we aim to alleviate this problem through feature space calibration. Specifically, we propose a novel loss function that allows the network to focus more on inter- and intra-class information to extract effective features. The balance between remembering old classes and learning new classes is achieved by learning class relationships rather than just information about a particular class, which can effectively alleviate catastrophic forgetting. Unlike existing methods, the approach proposed in this paper is highly general and flexible and can be directly integrated into existing CIL methods to effectively improve their performance. Our proposed approach is shown to be effective through comparative experiments on three popular datasets: CIFAR100, ImageNet100, and ImageNet1k. To ensure a robust comparison, we utilized three state-of-the-art methods as our baseline models. The results of these experiments demonstrate that our approach outperforms the baseline models on a range of benchmark datasets, showcasing its superiority and potential for broader application.

Zero-shot learning (ZSL) is an important and rapidly growing area of machine learning that aims to recognize new classes without prior training data. Despite its significance, ZSL has faced challenges with overfitting in embedding-based methods and limitations in traditional one-directional attention (ODA) based approaches. To bridge these gaps, this paper proposes the use of bi-directional attention (BDA) to integrate insights from both embedding and attention-based approaches. The proposed BDA system consists of a bi-directional attention network (BDAN) and a synthesized visual embedding network (SVEN) that facilitates visual-semantic interaction for ZSL classification. More specifically, the BDAN employs region self-attention (RSA), semantic synthesis attention (SSA), and visual synthesis attention (VSA) to overcome the overfitting issue in embedding methods and enhance transferability, to associate visual features with semantic property information, and to learn locally improved visual features. Extensive testing on CUB, SUN, and AWA2 datasets confirm the superiority of our proposed method over traditional approaches. Code is available at https://github.com/JunseokLee3/BDA.