Incremental Concept Learning via Online Generative Memory Recall.

Deep neural networks can be used to perform nonlinear operations at multiple levels, such as a neural network that is composed of many hidden layers. Although deep learning approaches show good results, they have a drawback called catastrophic forgetting, which is a reduction in performance when a new class is added. Incremental learning is a learning method where existing knowledge should be retained even when new data is acquired. It involves learning with multiple batches of training data and the newer learning sessions do not require the data used in the previous iterations. The Bayesian approach to incremental learning uses the concept of the probability distribution of weights. The key idea of Bayes theorem is to find an updated distribution of weights and biases. In the Bayesian framework, the beliefs can be updated iteratively as the new data comes in. Bayesian framework allows to update the beliefs iteratively in real-time as data comes in. The Bayesian model for incremental learning showed an accuracy of 82%. The execution time for the Bayesian model was lesser on GPU (670 s) when compared to CPU (1165 s).

With the advent of parameter-efficient fine-tuning techniques for pre-trained vision-language models, interest in adapting them for various incremental learning scenarios has grown, i.e., sequential increments on task, class, and domain. However, no high-performance incremental learning framework has integrated these three incremental scenarios to achieve Unified Incremental Learning (Uni-IL) in complex settings. In this work, we propose an incremental learning framework called Mixture of Attribute-Guided Experts (MAGE) to alleviate the long-term forgetting in vision-language model incremental learning. Our approach involves acquiring image attribute knowledge via LLMs to form an attribute pool. We match the most relevant attributes as inputs to the Mixture of Experts (MoE) to fine-tune the pre-trained CLIP. Then the expert routers learn to select specific expert combinations based on the data and attribute features, alleviating catastrophic forgetting. The attribute pool incorporates both domain and class knowledge, enabling our approach to adapt to the three types of incremental learning scenarios and thus facilitating unified incremental learning. Through extensive experiments on our newly proposed benchmark and existing incremental learning scenarios, the results demonstrate that our proposed method not only performs well on the new Uni-IL tasks but also consistently outperforms previous state-of-the-art methods. Source code is available at https://github.com/ElectricField/Uni-IL.

Contrastive Learning (CL) has been proved to be a powerful self-supervised approach for a wide range of domains, including computer vision and graph representation learning. However, the incremental learning issue of CL has rarely been studied, which brings the limitation in applying it to real-world applications. Contrastive learning identifies the samples with the negative ones from the noise distribution that changes in the incremental scenarios. Therefore, only fitting the change of data without noise distribution causes bias, and directly retraining results in low efficiency. To bridge this research gap, we propose a self-supervised Incremental Contrastive Learning (ICL) framework consisting of (i) a novel Incremental InfoNCE (NCE-II) loss function by estimating the change of noise distribution for old data to guarantee no bias with respect to the retraining, (ii) a meta-optimization with deep reinforced Learning Rate Learning (LRL) mechanism which can adaptively learn the learning rate according to the status of the training processes and achieve fast convergence which is critical for incremental learning. Theoretically, the proposed ICL is equivalent to retraining, which is based on solid mathematical derivation. In practice, extensive experiments in different domains demonstrate that, without retraining a new model, ICL achieves up to 16.7x training speedup and 16.8x faster convergence with competitive results.

Incremental learning empowers models to continuously acquire knowledge of new classes while retaining previously learned information. However, catastrophic forgetting and class imbalance often impede this process, especially when new classes are introduced sequentially. We propose a hybrid method that integrates Elastic Weight Consolidation (EWC) with a shared encoder architecture to overcome these obstacles. This approach provides robust feature extraction, while EWC safeguards vital parameters and preserves prior knowledge. Moreover, task-specific output layers enable flexible adaptation to new classes. We evaluated our method using the CICIoT2023 dataset, a class-incremental IoT anomaly detection benchmark. Our results demonstrated a 15.3% improvement in the macro F1-score and a 1.28% increase in overall accuracy compared to a baseline model that did not incorporate EWC, with particular advantages for underrepresented classes. These findings underscore the effectiveness of the EWC-assisted shared encoder framework for class-imbalanced incremental learning in streaming environments.

Deep neural networks are a branch in machine learning that has seen a meteoric rise in popularity due to its powerful abilities to represent and model high-level abstractions in highly complex data. One area in deep neural networks that are ripe for exploration is neural connectivity formation. A pivotal study on the brain tissue of rats found that synaptic formation for specific functional connectivity in neocortical neural microcircuits can be surprisingly well modeled and predicted as a random formation. Motivated by this intriguing finding, we introduce the concept of StochasticNet where deep neural networks are formed via stochastic connectivity between neurons. As a result, any type of deep neural networks can be formed as a StochasticNet by allowing the neuron connectivity to be stochastic. Stochastic synaptic formations in a deep neural network architecture can allow for efficient utilization of neurons for performing specific tasks. To evaluate the feasibility of such a deep neural network architecture, we train a StochasticNet using four different image datasets (CIFAR-10, MNIST, SVHN, and STL-10). Experimental results show that a StochasticNet using less than half the number of neural connections as a conventional deep neural network achieves comparable accuracy and reduces overfitting on the CIFAR-10, MNIST, and SVHN data sets. Interestingly, StochasticNet with less than half the number of neural connections, achieved a higher accuracy (relative improvement in test error rate of $\sim 6$ % compared to ConvNet) on the STL-10 data set than a conventional deep neural network. Finally, the StochasticNets have faster operational speeds while achieving better or similar accuracy performances.

Catastrophic forgetting is a major problem that affects neural networks during progressive learning. In it, the previously learned representation vanishes as the network learns new information. The extreme learning machine is one of the variants of the neural network. It is used in many domains due to fast training and good generalization ability. However, like other neural networks, it suffers from catastrophic forgetting and negative forward and backward transfer during the progression of neurons in incremental learning. The study hypothesizes that it is due to overlapping in hidden neurons and output weights. The global representation by an activation function further supports this hypothesis. To address this, the study proposes a neuron clustering approach to mitigate it in an adaptive class incremental extreme learning machine. The neuron clustering method activates k nearest neurons during learning and testing. It helps to partition the network to select overlapping subnetwork. Experimental results on four food datasets show that the proposed approach reduces negative forward and backward transfer when neurons are added incrementally during progressive learning.

Artificial neural networks have been reported to exhibit, and in some cases surpass, human level performance on individual rigid tasks. However, these networks remain static entities of knowledge for those specific tasks, which can lead to catastrophic forgetting ---i.e., forgetting old tasks--- when attempting to learn new tasks. The main objective of Incremental Learning (IL) is to address this issue. In order to tackle catastrophic forgetting, various approaches have been proposed so far. We survey those approaches and organize them in nine categories: regularization-based methods, exemplar replay-based methods, variational continual learning-based methods, parameter isolation-based methods, dynamic architectures-based methods, distillation-based methods, generative methods, data-free methods and unsupervised methods. Moreover, this review distinguishes between two scenarios, Task Incremental Learning (Task-IL) and Class Incremental Learning (Class-IL), and reports on the results obtained for a number of experiments that compare the performance achieved by a selection of diverse methods for both scenarios on the datasets most used by the related research community.

With the continuous research of deep neural networks, the network structure becomes more and more flexible, and it has achieved good results in terms of recognition, classification and so on. Traditional deep neural networks use one-time batch learning and suffer catastrophic forgetting when learning new data. This paper proposes a new fault diagnosis algorithm based on incremental structure. Through the use of improved residual neural networks, enhanced training sets, and enhanced loss functions, new knowledge can be continuously learned from new data while retaining what has been learned, thereby slowing catastrophic forgetting. Experiment demonstrates the effectiveness of the algorithm for mitigating catastrophic forgetting by using the Tennessee-Eastman (TE) chemical process dataset.

Many deep learning applications, like keyword spotting, require the incorporation of new concepts (classes) over time, referred to as Class Incremental Learning (CIL). The major challenge in CIL is catastrophic forgetting, i.e., preserving as much of the old knowledge as possible while learning new tasks. Various techniques, such as regularization, knowledge distillation, and the use of exemplars, have been proposed to resolve this issue. However, prior works primarily focus on the incremental learning step, while ignoring the optimization during the base model training. We hypothesize that a more transferable and generalizable feature representation from the base model would be beneficial to incremental learning. In this work, we adopt multitask learning during base model training to improve the feature generalizability. Specifically, instead of training a single model with all the base classes, we decompose the base classes into multiple subsets and regard each of them as a task. These tasks are trained concurrently and a shared feature extractor is obtained for incremental learning. We evaluate our approach on two datasets under various configurations. The results show that our approach enhances the average incremental learning accuracy by up to 5.5%, which enables more reliable and accurate keyword spotting over time. Moreover, the proposed approach can be combined with many existing techniques and provides additional performance gain.

Catastrophic forgetting in neural networks during incremental learning remains a challenging problem. Previous research investigated the catastrophic forgetting in fully connected networks with some earlier work exploring activation functions and learning algorithms. Applications of neural networks have been extended to include similarity and metric learning. It is of significant interest to understand how metric learning loss functions would be affected by catastrophic forgetting. Our research investigates catastrophic forgetting for four well-known metric-based loss functions during incremental class learning. The loss functions are angular, contrastive, center, and triplet loss. Our results show that the rate of forgetting is different across loss functions on multiple datasets. Triplet loss was least affected followed by contrastive, center, and angular loss. Center and angular loss produce better embeddings on difficult tasks when trained on all available training data, however, they are the least robust to forgetting during incremental class learning. We argue that triplet loss provides the ideal middle ground for future improvements.

Recently, the human-like behavior on anthropomorphic robot manipulators are increasingly accomplished by the kinematic model estabilshing the relationship of an anthropomorphic manipulator and human arm motions. Notably, the growth and broad availability of advanced techniques in data science facilitate the imitation learning process in anthropomorphic robotics. However, the enormous data set causes the labeling and prediction burden. In this paper, the swivel motion reconstruction approach was applied to imitate human-like behavior using the kinematic mapping in robot redundancy. For the sake of efficient computing, a novel incremental learning framework that combines an incremental learning approach with a deep convolutional neural network (IN-DCNN) is proposed for fast and efficient learning. The algorithm exploits a novel approach to detect changes from human motion data streaming and then to evolve its hierarchical representation of features. The incremental learning process is capable of fine-tuning the deep network only when model drifts detection mechanisms are triggered. Finally, we experimentally demonstrated this neural network's learning procedure and translated the trained human-like model to manage the redundancy optimization control of an anthropomorphic robot manipulator (LWR4+, KUKA, Germany). The anthropomorphic kinematic structure based redundant robots can be held by this approach. The experimental results showed that our architecture could not only enhance the regression accuracy but also significantly reduce the processing time of learning human motion data.

Transfer learning has shown promising results in leveraging loosely labeled Web images (source domain) to learn a robust classifier for the unlabeled consumer videos (target domain). Existing transfer learning methods typically apply source domain data to learn a fixed model for predicting target domain data once and for all, ignoring rapidly updating Web data and continuously changes of users requirements. We propose an incremental transfer learning framework, in which heterogeneous knowledge are integrated and incrementally added to update the target classifier during learning process. Under the framework, images (image source domain) queried from Web image search engine and videos (video source domain) from existing action datasets are adopted to provide static information and motion information of the target video, respectively. For the image source domain, images are partitioned into several groups according to their semantic information. And for the video source domain, videos are divided in the same way. Unlike traditional methods which measure relevance between the source group and the whole target domain videos, the group weights in this paper are treated as latent variables for each target domain video and learned automatically according to the probability distribution difference between the individual source group and target domain videos. Experimental results on the two challenging video datasets (i.e., CCV and Kodak) demonstrate the effectiveness of our proposed method.KeywordsTarget DomainGroup WeightTarget ClassifierTransfer LearningSource DomainThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Class-incremental learning (CIL) in deep neural networks is affected by catastrophic forgetting (CF), where acquiring knowledge of new classes leads to the significant degradation of previously learned representations. This challenge is particularly severe in medical image analysis, where costly, expertise-dependent annotations frequently contain pervasive and hard-to-detect noisy labels that substantially compromise model performance. While existing approaches have predominantly addressed CF and noisy labels as separate problems, their combined effects remain largely unexplored. To address this critical gap, this paper presents a dual-stage clean-sample selection method for Incremental Noisy Label Learning (DSCNL). Our approach comprises two key components: (1) a dual-stage clean-sample selection module that identifies and leverages high-confidence samples to guide the learning of reliable representations while mitigating noise propagation during training, and (2) an experience soft-replay strategy for memory rehearsal to improve the model's robustness and generalization in the presence of historical noisy labels. This integrated framework effectively suppresses the adverse influence of noisy labels while simultaneously alleviating catastrophic forgetting. Extensive evaluations on public medical image datasets demonstrate that DSCNL consistently outperforms state-of-the-art CIL methods across diverse classification tasks. The proposed method boosts the average accuracy by 55% and 31% compared with baseline methods on datasets with different noise levels, and achieves an average noise reduction rate of 73% under original noise conditions, highlighting its effectiveness and applicability in real-world medical imaging scenarios.

Meta-learning approaches have recently achieved promising performance in multi-class incremental learning. However, meta-learners still suffer from catastrophic forgetting, i.e., they tend to forget the learned knowledge from the old tasks when they focus on rapidly adapting to the new classes of the current task. To solve this problem, we propose a novel distilled meta-learning (DML) framework for multi-class incremental learning that integrates seamlessly meta-learning with knowledge distillation in each incremental stage. Specifically, during inner-loop training, knowledge distillation is incorporated into the DML to overcome catastrophic forgetting. During outer-loop training, a meta-update rule is designed for the meta-learner to learn across tasks and quickly adapt to new tasks. By virtue of the bilevel optimization, our model is encouraged to reach a balance between the retention of old knowledge and the learning of new knowledge. Experimental results on four benchmark datasets demonstrate the effectiveness of our proposal and show that our method significantly outperforms other state-of-the-art incremental learning methods.

In the field of machine learning, incremental learning methods allow models to learn new tasks without forgetting the old ones, but the catastrophic forgetting problem often encountered in this process greatly limits their applications. In this paper, an improved algorithm EPASS (Enhanced Prototype Augmentation and Self-Supervision) is proposed for incremental learning in image classification tasks, which effectively alleviates the limitations of the traditional PASS algorithm and the EWC algorithm in terms of catastrophic forgetting, sensitivity of parameter selection, and computational cost. The EPASS algorithm improves the flexibility and adaptability of the model by automatically selecting which prototypes to keep for different tasks through a dynamic prototype storage strategy. Meanwhile, the EWC-based prototype update strategy allows prototypes to be updated as new data arrives, further improving the model's adaptability to new tasks while reducing computational cost. Experimental results on the CIFAR-100 dataset show that the EPASS algorithm outperforms the PASS algorithm to some extent in terms of accuracy at all stages. These results validate the effectiveness of the EPASS algorithm in mitigating the catastrophic forgetting problem and provide new directions for future incremental learning research.