Few-shot class-incremental learning (FSCIL) requires learning new knowledge without forgetting old knowledge. Forward compatibility can reserve space for novel classes while maintaining base class knowledge in incremental learning. Better forward compatibility is crucial for effectively mastering all knowledge, especially when dealing with a few unknown new classes. In this article, we propose the better forward compatibility pretraining (BFCP) to further enhance forward compatibility in FSCIL. We adopt a two-stage training for the backbone network in the base session. First, we train the backbone network at the image-level to enhance its feature extraction capability, enabling the model to extract valuable information from unknown class images. Second, we fine-tune the backbone network at the feature-level with fake prototypes and instances to achieve clustering base classes and reserve space for unknown new classes. For all incremental new sessions, we freeze the backbone network and employ prototype rectification without further training to refine the prototypes of the novel classes. We conduct extensive experiments with different input scales, including federated cross-domain pretraining and cross-domain class-incremental experiments. BFCP efficiently handles both novel and base classes of each incremental session and significantly outperforms state-of-the-art methods, achieving an average accuracy of 63.47% on the CIFAR100 dataset.

Few-shot learning (FSL) is challenging due to the scarce labeled novel-class data. Researchers have to train the embedding function with auxiliary base-class data to obtain the novel-class embeddings. However, the domain gap makes the novel-class embedding unsatisfactory, as the novel class and the base class are disjoint. Recent studies prove that embedding rectification shows great potential, introduces miscellaneous variants, and achieves similar performances. Nonetheless, while each method demonstrates unique strengths, they often address distinct challenges in isolation, limiting their applicability in more complex or diverse scenarios. In this article, we take a closer look at these methods and hypothesize that a general embedding rectification framework is more essential to the model's performance. To verify our observation, we propose: 1) a distribution propagation (DisP) layer distinguishes the inter-class margin and increases intra-class aggregation, performing the task-level rectification; and 2) a prototype propagation (ProtoP) layer moves the prototype toward the ideal class center, applying the prototype-query level rectification. Our framework aims to maximize the actual data distribution. Although pseudo-labeling proves effective in achieving this goal, a significant challenge is ensuring the reliable retention of only high-confidence predictions. To overcome this, we introduce a distribution-based pseudo-labeling method pseudo-query upgrade (PseQUp) that provides more reliable pseudo-labeling samples without relying on confidence scores. We evaluate the proposed method in both transfer learning and meta-learning scenarios. Empirical experiments show the applicable and plug-and-play ability of the proposed methods.

Sparse subspace clustering algorithm cluster the data points located on the union of low-dimensional subspaces through the ℒ1minimization program. However, the ℒ1-norm is not rotation invariant, and utilizing original data containing noise as the dictionary leads to poor performance. This paper proposes a method named robust structured sparse subspace clustering with neighborhood preserving projection (RSSSC). Firstly, RSSSC replaces the ℒ1minimization program with a structured re-weighting sparse regularization term, effectively recovering the sparse representation. Secondly, RSSSC uses the extracted features as the dictionary in the self-representation problem to replace the original dataset containing noise and outliers, thus making the model more robust. By fully considering the low-dimensional manifold structure of samples in the original high-dimensional space, RSSSC preserves the neighborhood structure while learning the optimal projection. We verify the effectiveness of the proposed method through experiments on real-world image datasets.

In few-shot learning, methods are enslaved to the scarce labeled data, resulting in suboptimal embedding. Recent studies learn the embedding network by other large-scale labeled data. However, the trained network may give rise to the distorted embedding of target data. We argue two respects are required for an unprecedented and promising solution. We call them Better Embedding and More Shots (BEMS). Suppose we propose to extract embedding from the embedding network. BE maximizes the extraction of general representation and prevents over-fitting information. For this purpose, we introduce the topological relation for global reconstruction, avoiding excessive memorizing. MS maximizes the relevance between the reconstructed embedding and the target class space. In this respect, increasing the number of shots is a pivotal but intractable strategy. As a creative method, we derive the bound of information-theory-based loss function and implicitly achieve infinite shots with negligible cost. A substantial experimental analysis is carried out to demonstrate the state-of-the-art performance. Compared to the baseline, our method improves by up to 10%+. We also prove that BEMS is suitable for both standard pre-trained and meta-learning embedded networks.

Semisupervised multiview learning gains extensive research attention due to its strong capability to utilize the heterogeneous features and the label information of a few labeled samples. However, the supervision information is not well utilized in the process of exploring the consensus structure of the multiview data. In this paper, we propose a novel unified pseudolabel-guided multiview consensus (PMvC) learning framework for the semisupervised classification problem, which learns the consensus structure of multiview data by fully exploiting the supervised information of labeled samples. Specifically, PMvC first assigns multiple pseudolabels to the unlabeled samples by selecting the nearest labeled sample in each view separately, and then labels the part of unlabeled samples by selecting the pseudolabel that agrees across all views. By doing so, the high-confident pseudolabeled samples can be selected to enlarge the labeled sample pool and the supervision information can be exploited further in the learning process. In addition, to capture the consensus structure of the multiview data, PMvC learns a consensus graph from the view-specific self-representation graph guided by enhanced supervision information, which better preserves the manifold structure of samples. Meanwhile, the label information is also propagated from the labeled samples to the unlabeled samples by the learned consensus graph simultaneously. Accordingly, an effective optimization algorithm is derived to find the optimal solution for PMvC. Extensive experiment results on several real-world data sets demonstrate the feasibility and superiority of PMvC. The source code of PMvC is available at https://github.com/justcallmewilliam/PMvC.

Multiple empirical kernel learning (MEKL) is a scalable and efficient supervised algorithm based on labeled samples. However, there is still a huge amount of unlabeled samples in the real-world application, which are not applicable for the supervised algorithm. To fully utilize the spatial distribution information of the unlabeled samples, this paper proposes a novel semi-supervised multiple empirical kernel learning (SSMEKL). SSMEKL enables multiple empirical kernel learning to achieve better classification performance with a small number of labeled samples and a large number of unlabeled samples. First, SSMEKL uses the collaborative information of multiple kernels to provide a pseudo labels to some unlabeled samples in the optimization process of the model, and SSMEKL designs pseudo-empirical loss to transform learning process of the unlabeled samples into supervised learning. Second, SSMEKL designs the similarity regularization for unlabeled samples to make full use of the spatial information of unlabeled samples. It is required that the output of unlabeled samples should be similar to the neighboring labeled samples to improve the classification performance of the model. The proposed SSMEKL can improve the performance of the classifier by using a small number of labeled samples and numerous unlabeled samples to improve the classification performance of MEKL. In the experiment, the results on four real-world data sets and two multiview data sets validate the effectiveness and superiority of the proposed SSMEKL.