Search

Significant progress has been made in the semantic segmentation of mobile laser scanning (MLS) point clouds based on deep learning. However, the segmentation classes of deep learning models depend on the label classes of the source point clouds used for training, which makes it difficult to generalize the models to target point clouds with novel classes. In addition, retraining models using complete class label datasets is time-consuming, and the source point clouds are often unavailable or occupy a large amount of storage space. In this paper, we propose a new class-incremental semantic segmentation framework for MLS point clouds. Firstly, to prevent catastrophic forgetting of original class knowledge when the model learns novel classes, we design a feature representation preservation-based knowledge distillation module to maintain the encoding ability of the target models for original classes. Then, to further separate novel classes from the original background classes, we introduce a background shift mechanism based on loss cross-coupling and pseudo-label collaborative training, which adaptively balances the model plasticity when learning novel class knowledge. Finally, we conducted extensive experiments on two benchmark datasets (Paris-Lille-3D and Toronto-3D), and our proposed method achieved impressive results, which indicate that the proposed framework could effectively achieve class-incremental semantic segmentation for MLS point clouds. [ABSTRACT FROM AUTHOR]

Federated Learning has received much attention due to its data privacy benefits, but most existing approaches assume that client classes are fixed. Clients may remove old classes and add new ones, leading to catastrophic forgetting of the model. Existing methods have limitations, such as requiring additional client storage and distillation methods becoming less effective as new classes increase. For this reason, this paper proposes the Data-Model Anti-Forgetting (DMAF) framework. Specifically, in the proposed framework, an auxiliary client and group aggregation method to mitigate catastrophic forgetting at the data level has been proposed, which does not require clients to allocate additional storage space to store synthetic data and can balance class distributions. A multi-teacher integrated knowledge distillation method was adopted to retain old class knowledge by distilling multiple teacher models and design task fusion for further tuning of the global model. Finally, this paper conducts extensive experiments on public datasets to validate the effectiveness of DMAF. [ABSTRACT FROM AUTHOR]

Incremental learning is a machine learning strategy that enables the integration of new data to address emerging tasks without retraining the model from scratch. This approach retains previously acquired knowledge and conserves resources, yet it faces the challenge of catastrophic forgetting. Hierarchical classification (HC) improves accuracy and efficiency by assigning labels with hierarchical relationships to objects. In this work, we propose CHFL, an incremental learning method for remote sensing images. CHFL leverages hierarchical relationships through a label relationship graph and class hierarchy information to encode knowledge effectively. It integrates a Feature Pyramid Network (FPN) to process multi-scale features, capturing discriminative region information, and a Learning Without Forgetting (LWF) strategy to efficiently learn new classes while preserving performance on previous ones. Additionally, a Hierarchy and Exclusion (HEX) graph is introduced to constrain label predictions, enhancing consistency and improving classification accuracy. Experimental results on the high-resolution remote sensing dataset HRSC show that CHFL achieves an accuracy of 93.82% on new class classifications while maintaining competitive performance on previous classes. Compared with existing methods, CHFL demonstrates superior classification performance, effectively mitigating catastrophic forgetting and addressing scale variations in remote sensing imagery. [ABSTRACT FROM AUTHOR]

An adaptive machine learning framework is established for an implicit determination of the performance degradation of a ship due to marine growth, i.e. biofouling. The framework is applied in a case study considering telemetry data of a cruise ship operating predominantly in the Caribbean Sea. The dataset encompasses seven years including three dry-docking intervals and several in-water cleaning events. The COVID-19 period receives special focus due to the drastic change in the operational profile. A main outcome of the study is a comparison of the derived performance estimate to the corresponding results of the industry standard ISO 19030. Additional aspects of the present study include the use of special regularization techniques for incremental machine learning and the increase of transparency through the implementation of prediction intervals indicating model uncertainty. Overall, it is found that the developed machine learning framework shows good agreement with the industry standard underlining its plausibility. [ABSTRACT FROM AUTHOR]

Event detection is a crucial task in natural language processing, and it plays a significant role in numerous applications, such as information retrieval, question answering, and situational awareness. Real-world tasks typically require robust models that can dynamically adapt to changing data distributions and seamlessly accommodate emerging event types while maintaining high accuracy and efficiency. However, existing methods often face catastrophic forgetting, a significant challenge where models lose previously acquired knowledge when learning new information. This phenomenon hinders models from balancing performance with adaptability, limiting their ability to generalize across dynamic data landscapes. This paper proposes a novel event detection framework, DASR, which aims to enhance the flexibility and diversity of event detection through joint learning and guidance that dynamically adapts to new events and extracts semantic relevance. Firstly, we utilize pre-trained language models (PLMs) trained on general corpora to obtain existing event and type information as global knowledge. Secondly, during prompt fine-tuning for specific tasks, we incorporate an incremental learning module to design incremental prompt templates for newly introduced event types and read out their representations within the PLM. Subsequently, we perform entity recognition and event trigger word detection to extract semantic relevance. In this case, a graph attention mechanism is introduced to enhance the long-distance dependencies within the text (modeled as message passing in the type graph). Additionally, feature fusion integrates entity and event trigger word information into a unified representation. Finally, we validate the effectiveness of the proposed framework through extensive experiments. The experimental results demonstrate that the proposed framework effectively mitigates catastrophic forgetting and significantly improves the accuracy and adaptability of event detection when dealing with evolving data distributions and newly introduced event types. [ABSTRACT FROM AUTHOR]

Catastrophic forgetting in neural networks during incremental learning remains a challenging problem. Previous research investigated 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 learning. Understanding how similarity learning loss functions would be affected by catastrophic forgetting is of significant interest. Our research investigates catastrophic forgetting for four well-known similarity-based loss functions during incremental class learning. The loss functions are Angular, Contrastive, Center, and Triplet loss. Our results show that the catastrophic forgetting rate differs across loss functions on multiple datasets. The Angular loss was least affected, followed by Contrastive, Triplet loss, and Center loss with good mining techniques. We implemented three existing incremental learning techniques, iCaRL, EWC, and EBLL. We further proposed a novel technique using Variational Autoencoders (VAEs) to generate representation as exemplars passed through the network's intermediate layers. Our method outperformed three existing state-of-the-art techniques. We show that one does not require stored images (exemplars) for incremental learning with similarity learning. The generated representations from VAEs help preserve regions of the embedding space used by prior knowledge so that new knowledge does not "overwrite" it. [ABSTRACT FROM AUTHOR]

Training data of many vision tasks may be sequentially arrived in practice, e.g., the vision tasks in autonomous driving or video surveillance applications. This raises a fundamental challenge that, how to keep improving the performance on a specific task by learning from sequentially available training splits. This paper investigates this task as Incremental Model Enhancement (IME). IME is distinct from the conventional Incremental Learning (IL), where each training split typically corresponds to a set of independent classes, domains, or tasks. In IME, each training split may only cover part of the entire data distribution for the target vision task. Consequently, the IME model should be optimized towards the joint distribution of all available training splits, instead of optimizing towards each newly arrived one like IL methods. To deal with above issues, our method stores feature vectors of previously observed training data in the memory bank, which preserves compressed knowledge of the previous training data. We hence adopt the memorized features and each newly arrived training split for training via Memory-based Contrastive Learning (MCL). A new Contrastive Relation Preserving (CRP) scheme updates the memory bank to prevent obsoleteness of the preserved features and works with MCL simultaneously to boost the model performance. Experiments on several large-scale image classification benchmarks demonstrate the effectiveness of our method. Our method also works well on semantic segmentation, showing strong generalization ability on diverse vision tasks. [ABSTRACT FROM AUTHOR]

Domain-specific vocabulary, which is crucial in fields such as Information Retrieval and Natural Language Processing, requires continuous updates to remain effective. Incremental Learning, unlike conventional methods, updates existing knowledge without retraining from scratch. This paper presents an incremental learning algorithm for updating domain-specific vocabularies. It introduces DocLib, an archive used to capture a compact footprint of previously seen data and vocabulary terms. Task-based evaluation measures the effectiveness of the updated vocabulary by using vocabulary terms to perform a downstream task of text classification. The classification accuracy gauges the effectiveness of the vocabulary in discerning unseen documents related to the domain. Experiments illustrate that multiple incremental updates maintain vocabulary relevance without compromising its effectiveness. The proposed algorithm ensures bounded memory and processing requirements, distinguishing it from conventional approaches. Novel algorithms are introduced to assess the stability and plasticity of the proposed approach, demonstrating its ability to assimilate new knowledge while retaining old insights. The generalizability of the vocabulary is tested across datasets, achieving 97.89% accuracy in identifying domain-related data. A comparison with state-of-the-art techniques using a benchmark dataset confirms the effectiveness of the proposed approach. Importantly, this approach extends beyond classification tasks, potentially benefiting other research fields. [ABSTRACT FROM AUTHOR]

The exponential growth of digital networks necessitates robust intrusion detection systems (IDS) to counter evolving cyber threats effectively. Machine learning offers adaptive solutions for these challenges. This study evaluates the comparative performance of Incremental Learning and Batch Learning methodologies for IDS using two datasets, UNSW-NB15 and CI-CIDS 2017. Three machine learning algorithms—Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF)—were assessed. Results indicate that Incremental Learning outperforms Batch Learning in dynamic environments. For instance, on the CI-CIDS 2017 dataset, SVM with Incremental Learning achieved a precision of 98%, recall of 97%, and an F1-score of 97.5%, compared to Batch Learning, which obtained 95%, 93%, and 94%, respectively. These findings highlight Incremental Learning's adaptability to real-time threats despite higher computational demands. This research offers valuable insights for optimizing IDS using machine learning and proposes a framework for enhancing network security. [ABSTRACT FROM AUTHOR]

With the continuous expansion of data from different domains, there are lots of inconsistent data in decision tables. Approximations are fundamental notions of rough set theory. However, inconsistent decision tables (IDTs) may update with new coming objects due to the information collection and update. In this circumstance, how to dynamically acquire useful approximations with previous knowledge in IDTs is crucial for data analysis and knowledge discovery. To address this issue, we focus on exploiting incremental approaches for computing approximations when objects are changing in IDTs over time, which is of great significance for gaining decision rules in the context of dynamic environments. First, we introduce the concepts of inconsistent decision tables and approximations. Furthermore, with dynamic changing objects, several principles of updating approximations are investigated according to the positive and boundary regions. Following the proposed principles, incremental algorithms for acquiring approximations in IDTs are presented while the objects change with time. In the end, comparative results are illustrated to verify effectiveness of updating lower and upper approximations in IDTs, displaying the advantage of the proposed approaches.Article Highlights: Incremental principles of obtaining approximations with the changes of objects in IDTs are presented. In line with the principles of knowledge maintenance, the incremental algorithms are designed when adding an object or deleting an object in IDTs. Experimental results on public data sets show that the proposed methods are effective regarding runtime when updating approximations. [ABSTRACT FROM AUTHOR]

As the demand for tunnel inspection grows, visual methods are being utilized, but traditional machine learning techniques struggle with continuous learning and updating prototype classes without repeated defect data acquisition. Incremental learning provides a way to address catastrophic forgetting, but conventional methods often cannot handle the imbalanced data typical in real-world situations. This paper proposes a two-stage learning paradigm specifically designed for incremental learning to reduce overfitting from head data in unbalanced classes. Our approach maximizes symmetric separation of inter-class prototypes in the classifier space by integrating DR loss within the ETF classifier. Compared to other leading incremental learning methods, our approach shows superior performance on tunneling data. [ABSTRACT FROM AUTHOR]

Human pose estimation has a wide range of applications. Existing methods perform well in conventional domains, but there are certain defects when they are applied to sports activities. The first is lack of estimation of the extremity posture, making it impossible to comprehensively evaluate the movement posture; the second is insufficient occlusion handling. Therefore, we propose a human pose compensation network based on incremental learning, which obtains shared weights to extract detailed features under the premise of limited extremity training data. We propose a higher-order feature compensator (HOF-compensator) to embed the attributes of the extremity into the torso and limbs topology structure, building a complete higher-order feature. In addition, to improve the occlusion handling performance, we propose an occlusion feature enhancement attention mechanism (OFE-attention) that can identify occluded keypoints and enhance attention to occlusion areas. We design comparative experiments on three public datasets and a self-built sports dataset, achieving the highest mean accuracy among all comparative methods. In addition, we design a series of ablation analysis and visualization displays to verify that our method performs best in sports pose estimation. [ABSTRACT FROM AUTHOR]

Abstract With the development of data acquisition technology, a large amount of time-series data can be collected. However, handling too much data often leads to a waste of social resources. It becomes significant to determine the minimum data size required for training. In this paper, a framework for neural ordinary differential equations based on incremental learning is discussed, which can enhance learning ability and determine the minimum data size required in data modeling compared to neural ordinary differential equations. This framework continuously updates the neural ordinary differential equations with newly added data while avoiding the addition of extra parameters. Once the preset accuracy is reached, the minimum data size needed for training can be determined. Furthermore, the minimum data size required for five classic models under various sampling rates is discussed. By incorporating new data, it enhances accuracy instead of increasing the depth and width of the neural network. The close integration of data generation and training can significantly reduce the total time required. Theoretical analysis confirms convergence, while numerical results demonstrate that the framework offers superior predictive ability and reduced computation time compared to traditional neural differential equations.

In object detection task, incremental learning method enables the previously trained model better adapt to the new task using either a small amount of old data or none at all. In the incremental training processes of complex remote sensing scenes, the newly arrived data only includes new classes annotations. These new classes may exhibit spatial overlap and shape similarity with old classes or may have been labeled as background in earlier tasks, leading to a unique challenge called class semantic confusion. To address this issue, this article dynamically generate multiple representative prototypes of the different categories for refined matching the objects. To improve the matching accuracy, prototype contrastive learning is employed for expanding the distance between dissimilar prototypes and reducing the distance between similar prototypes. Meanwhile, a category perception enhancement module is proposed to enhance the aware of old categories to mitigate catastrophic forgetting. Comprehensive experimental results demonstrate that our proposed method outperforms the current state-of-the-art class-incremental object detection methods in most experimental settings on DIOR and FAIR1M datasets.

Deep learning techniques have achieved significant success in Synthetic Aperture Radar (SAR) target recognition using predefined datasets in static scenarios. However, real-world applications demand that models incrementally learn new information without forgetting previously acquired knowledge. The challenge of catastrophic forgetting, where models lose past knowledge when adapting to new tasks, remains a critical issue. In this paper, we introduce IncSAR, an incremental learning framework designed to tackle catastrophic forgetting in SAR target recognition. IncSAR combines the power of a Vision Transformer (ViT) and a custom-designed Convolutional Neural Network (CNN) in a dual-branch architecture, integrated via a late-fusion strategy. Additionally, we explore the use of TinyViT to reduce computational complexity and propose an attention mechanism to dynamically enhance feature representation. To mitigate the speckle noise inherent in SAR images, we employ a denoising module based on a neural network approximation of Robust Principal Component Analysis (RPCA), leveraging a simple neural network for efficient noise reduction in SAR imagery. Moreover, a random projection layer improves the linear separability of features, and a variant of Linear Discriminant Analysis (LDA) decorrelates extracted class prototypes for better generalization. Extensive experiments on the MSTAR, SAR-AIRcraft-1.0, and OpenSARShip benchmark datasets demonstrate that IncSAR significantly outperforms state-of-the-art approaches, achieving a 99.63% average accuracy and a 0.33% performance drop, representing an 89% improvement in retention compared to existing techniques. The source code is available at https://github.com/geokarant/IncSAR.

In the study of incremental object detection, knowledge distillation and data replay are effective methods to mitigate catastrophic forgetting. However, current research on single-stage detectors is limited, single-stage detector outputs often contain excessive negative sample information, and direct application of knowledge distillation to them is ineffective. To address this, this paper proposes a multi-level foreground prompt incremental learning algorithm for single-stage detectors like YOLO, including foreground prompts at the image level, feature map level, and knowledge level. First, to obtain fewer but more representative replay samples, representative images with a high number of old class foregrounds are selected by maximizing sample structure distance, providing direct foreground information at the image level. Second, the feature map output by the teacher model is used as a feature-level prompt, with a feature distillation loss guiding the student model to encode new class foreground information in less significant channels of the old feature map, reducing interference. Lastly, the teacher model’s inference output serves as a knowledge-level prompt, and the adaptive select object method is introduced to avoid foreground conflict in traditional knowledge distillation, enhancing the model’s plasticity by selectively merging foreground information. Extensive experiments on PASCAL VOC and MS COCO datasets demonstrate that this approach significantly improves model plasticity while maintaining stability.

Human-machine interfaces based on myoelectric signals typically use screen-guided training (SGT) for model calibration, but this approach fails to capture realistic user behaviors. This study evaluates a user-in-the-loop context-informed incremental learning (CIIL) framework, comparing SGT, SGT followed by CIIL adaptation (SGT-A), and a novel zero-shot adaptation (ZS-A) CIIL approach that begins adapting with no prior training. Sixteen participants completed a Fitts’ Law targeting task using these control schemes, with performance measured via online throughput and offline classification accuracy. Despite lower offline accuracy, the ZS-A model achieved the highest online throughput ( $1.47~\pm ~0.46$ bits/s), significantly outperforming the SGT baseline ( $1.15~\pm ~0.37$ bits/s) and reached competitive performance within 200 seconds. To further enhance control performance, a novel adaptive sigmoid-based proportional control mapping was introduced, dynamically adjusting control signals to allow precise control near neutral positions and rapid movements at higher activation levels, better aligning with natural user behaviors. These findings demonstrate that CIIL can surpass traditional SGT methods in online performance and emphasize the value of real-time user-in-the-loop data for developing adaptable and intuitive myoelectric interfaces, with implications for prosthetics, rehabilitation, and telerobotics.

This paper introduces Incremental Learning MPC (ILMPC), a novel Model Predictive Control (MPC) approach designed to enhance the adaptability of control systems in dynamic environments with unpredictable disturbances. Traditional MPC methods are often limited by their reliance on static models and fixed optimization schemes, making them less effective in handling disturbances and model inaccuracies. To overcome these limitations, ILMPC integrates incremental learning, enabling continuous refinement of the control model using real-time data. This innovation improves prediction accuracy and control performance, allowing the system to adapt to changing operational conditions and unknown disturbances. Key advances include the development of a sequence prediction model that continuously updates the state-space model through incremental learning, improved disturbance suppression for more stable control, and a reduction in computational complexity by incrementally model parameters. Experimental results show that ILMPC enhances deviation suppression significantly compared to conventional methods and significantly reduces control input volatility, demonstrating its superior performance in real-time disturbance suppression and adaptability. [ABSTRACT FROM AUTHOR]

Abstract With the continuous expansion of data from different domains, there are lots of inconsistent data in decision tables. Approximations are fundamental notions of rough set theory. However, inconsistent decision tables (IDTs) may update with new coming objects due to the information collection and update. In this circumstance, how to dynamically acquire useful approximations with previous knowledge in IDTs is crucial for data analysis and knowledge discovery. To address this issue, we focus on exploiting incremental approaches for computing approximations when objects are changing in IDTs over time, which is of great significance for gaining decision rules in the context of dynamic environments. First, we introduce the concepts of inconsistent decision tables and approximations. Furthermore, with dynamic changing objects, several principles of updating approximations are investigated according to the positive and boundary regions. Following the proposed principles, incremental algorithms for acquiring approximations in IDTs are presented while the objects change with time. In the end, comparative results are illustrated to verify effectiveness of updating lower and upper approximations in IDTs, displaying the advantage of the proposed approaches.

Abstract Human pose estimation has a wide range of applications. Existing methods perform well in conventional domains, but there are certain defects when they are applied to sports activities. The first is lack of estimation of the extremity posture, making it impossible to comprehensively evaluate the movement posture; the second is insufficient occlusion handling. Therefore, we propose a human pose compensation network based on incremental learning, which obtains shared weights to extract detailed features under the premise of limited extremity training data. We propose a higher-order feature compensator (HOF-compensator) to embed the attributes of the extremity into the torso and limbs topology structure, building a complete higher-order feature. In addition, to improve the occlusion handling performance, we propose an occlusion feature enhancement attention mechanism (OFE-attention) that can identify occluded keypoints and enhance attention to occlusion areas. We design comparative experiments on three public datasets and a self-built sports dataset, achieving the highest mean accuracy among all comparative methods. In addition, we design a series of ablation analysis and visualization displays to verify that our method performs best in sports pose estimation.

Global Positioning System (GPS)/Inertial Navigation System (INS) integration is a widely used technique for navigation and positioning applications. It combines the advantages of GPS and INS to provide accurate and reliable information. However, the GPS/INS integration suffers from performance degradation during a GPS outage, which occurs when natural or artificial factors block the GPS signal. The novelty of this paper is improving GPS/INS integration performance during GPS outages using Incremental Regularized LSTM (IncRLSTM) learning. Incremental learning is a learning paradigm that can be learned from streaming data online and updating the model parameters without forgetting the previous ones. Also, regularization is a technique that prevents overfitting and improves the generalization of the network by adding some constraints or penalties to the model. IncRLSTM learning models the GPS signal as a multi-objective regression and corrects the INS output with the Faded Memory Kalman filter. The results on real-world datasets significantly show the reduction of positioning errors by an average of 72 % during GPS outages and the improvement of the accuracy and robustness of GPS/INS integration by an average of 58 % compared with existing methods. Moreover, IncRLSTM presents, on average, a 10 % improvement compared to the existing methods.

Abstract Modern natural language processing (NLP) state-of-the-art (SoTA) deep learning (DL) models have hundreds of millions of parameters, making them extremely complex. Large datasets are required for training these models, and while pretraining has reduced this requirement, human-labelled datasets are still necessary for fine-tuning. Few-shot learning (FSL) techniques, such as meta-learning, try to train models from smaller datasets to mitigate this cost. However, the tasks used to evaluate these meta-learners frequently diverge from the problems in the real world that they are meant to resolve. This work aims to apply meta-learning to a problem that is more pertinent to the real world: class incremental learning (IL). In this scenario, after completing its training, the model learns to classify newly introduced classes. One unique quality of meta-learners is that they can generalise from a small sample size to classes that have never been seen before, which makes them especially useful for class incremental learning (IL). The method describes how to emulate class IL using proxy new classes. This method allows a meta-learner to complete the task without the need for retraining. To generate predictions, the transformer-based aggregation function in a meta-learner that modifies data from examples across all classes has been proposed. The principal contributions of the model include concurrently considering the entire support and query sets, and prioritising attention to crucial samples, such as the question, to increase the significance of its impact during inference. The outcomes demonstrate that the model surpasses prevailing benchmarks in the industry. Notably, most meta-learners demonstrate significant generalisation in the context of class IL even without specific training for this task. This paper establishes a high-performing baseline for subsequent transformer-based aggregation techniques, thereby emphasising the practical significance of meta-learners in class IL.

Abstract The outbreak of emerging infectious diseases poses significant challenges to global public health. Accurate early forecasting is crucial for effective resource allocation and emergency response planning. This study aims to develop a comprehensive predictive model for emerging infectious diseases, integrating the blending framework, transfer learning, incremental learning, and the biological feature Rt to increase prediction accuracy and practicality. By transferring features from a COVID-19 dataset to a monkeypox dataset and introducing dynamically updated incremental learning techniques, the model's predictive capability in data-scarce scenarios was significantly improved. The research findings demonstrate that the blending framework performs exceptionally well in short-term (7-day) predictions. Furthermore, the combination of transfer learning and incremental learning techniques significantly enhanced the adaptability and precision, with a 91.41% improvement in the RMSE and an 89.13% improvement in the MAE. In particular, the inclusion of the Rt feature enabled the model to more accurately reflect the dynamics of disease spread, further improving the RMSE by 1.91% and the MAE by 2.17%. This study underscores the significant application potential of multimodel fusion and real-time data updates in infectious disease prediction, offering new theoretical perspectives and technical support. This research not only enriches the theoretical foundation of infectious disease prediction models but also provides reliable technical support for public health emergency responses. Future research should continue to explore integrating data from multiple sources and enhancing model generalization capabilities to further enhance the practicality and reliability of predictive tools.

Effective monitoring of road conditions is crucial for ensuring safe and efficient transportation systems. By leveraging the power of crowd-sourced smartphone sensor data, road condition monitoring can be conducted in real-time, providing valuable insights for transportation planners, policymakers, and the general public. Previous studies have primarily focused on the use of pre-trained machine learning models and threshold-based methods for anomaly classification, which may not be suitable for real-world scenarios that require incremental detection and classification. As a result, there is a need for novel approaches that can adapt to changing data environments and perform effective classification without relying on pre-existing training data. This study introduces a novel, real-time road condition monitoring technique harnessing smartphone sensor data, addressing the limitations of pre-trained models that lack adaptability in dynamic environments. A hybrid anomaly detection method, combining unsupervised and supervised learning, is proposed to effectively manage concept drift, demonstrating a significant improvement in accuracy and robustness with a 96% success rate. The findings underscore the potential of incremental learning to enhance model responsiveness and efficiency in distinguishing various road anomalies, offering a promising direction for future transportation safety and resource optimization strategies. [ABSTRACT FROM AUTHOR]

Traditional object detection algorithms operate within a closed set, where the training data may not cover all real-world objects. Therefore, the issue of open-world object detection has attracted significant attention. Open-world object detection faces two major challenges: "neglecting unknown objects" and "misclassifying unknown objects as known ones." In our study, we address these challenges by utilizing the Region Proposal Network (RPN) outputs to identify potential unknown objects with high object scores that do not overlap with ground truth annotations. We introduce the reselection mechanism, which separates unknown objects from the background. Subsequently, we employ the simulated annealing algorithm to disentangle features of unknown and known classes, guiding the detector's learning process. Our method has improved on multiple evaluation metrics such as U-mAP, U-recall, and UDP, greatly alleviating the challenges faced by open world object detection. [ABSTRACT FROM AUTHOR]

The weld crack leakage due to stress concentration and external load is a significant safety risk in pressure pipelines. Microstructural variations and dynamic propagation lead to unpredictable changes in leakage rate over time and conditions. To address the above problems, a novel framework called OILS-TCN for weld crack pattern recognition and leakage rate prediction is proposed. Firstly, the adaptive threshold optimization algorithm is introduced into the self-organizing incremental neural network to update and increase the crack leakage pattern. Secondly, the depth first search algorithm is combined with the radial basis function neural network to perform online increment labelling of the leakage state. Then, according to the attenuation characteristics of acoustic emission signals, a portable input-attention module is designed to add different weights to the input sequence. Finally, the accurate prediction of leakage rate under different conditions is realized based on the temporal convolutional network. Compared with other advanced methods, the proposed method has obvious advantages in the adaptability and accuracy of pipeline weld crack leakage rate prediction. In addition, the validity and necessity of each part of the framework proposed are discussed based on ablation experiments. The proposed method can predict the leakage rate in real time without modifying the hyperparameters of the model, and can provide a powerful guide for the online monitoring of the leakage AE technology of pressure pipeline in complex systems. [ABSTRACT FROM AUTHOR]

The degradation of the pretightening state of bolts is a multistage process. Utilising a single model to monitor the pretightening state in the full degradation stage and continuously updating the monitoring model is important. Therefore, a quantitative monitoring method for the pretightening state of bolts based on nonlinear Lamb waves and incremental learning is proposed. In the proposed method, phase reversal technology is first adopted to enhance the sensitivity for bolt loosening, and then a relative nonlinear coefficient based on phase reversal (RCP) is constructed. The disadvantage that linear indicators are insensitive to early loosening is overcome and the critical points of the multidegradation process are identified by this indicator, the tight contact stage (TCS) and the significant loosening stage. After the TCS is determined, a quantitative monitoring model, which fuses seven nonlinear damage indexes, is established based on incremental canonical correlation forests (ICCF). This algorithm achieves incremental learning by continuously increasing the number of decision trees. To verify the effectiveness of the method, an experimental study is carried out on four bolts. The monitoring effects of different indicators show that the method has higher accuracy and retains the ability to dynamically update data. [ABSTRACT FROM AUTHOR]

Copyright of Experimental Technology & Management is the property of Experimental Technology & Management Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Gesture recognition techniques based on surface electromyography (sEMG) signals face instability problems caused by electrode displacement and the time-varying characteristics of the signals in cross-time applications. This study proposes an incremental learning framework based on densely connected convolutional networks (DenseNet) to capture non-synchronous data features and overcome catastrophic forgetting by constructing replay datasets that store data with different time spans and jointly participate in model training. The results show that, after multiple increments, the framework achieves an average recognition rate of 96.5% from eight subjects, which is significantly better than that of cross-day analysis. The density-based spatial clustering of applications with noise (DBSCAN) algorithm is utilized to select representative samples to update the replayed dataset, achieving a 93.7% recognition rate with fewer samples, which is better than the other three conventional sample selection methods. In addition, a comparison of full dataset training with incremental learning training demonstrates that the framework improves the recognition rate by nearly 1%, exhibits better recognition performance, significantly shortens the training time, reduces the cost of model updating and iteration, and is more suitable for practical applications. This study also investigates the use of the incremental learning of action classes, achieving an average recognition rate of 88.6%, which facilitates the supplementation of action types according to the demand, and further improves the application value of the action pattern recognition technology based on sEMG signals. [ABSTRACT FROM AUTHOR]

The identification of agricultural lands is crucial for sustainable development in rural areas. In this paper, an augmented model that utilizes multimodal satellite-based data samples for identifying agricultural lands via incremental learning. The Whale Optimization Algorithm (WOA) is used for augmenting collected images and data samples to enhance the accuracy of the model under real-time conditions. To identify agricultural lands, Deep Convolutional Neural Networks (CNNs) are trained on the augmented data samples. Additionally, the incorporation of Q-Learning for continuous optimization of the model to enhance its efficiency and effectiveness in identifying agricultural lands. The proposed model offers many edges over existing methods. Firstly, the use of multimodal satellite-based data samples allows for a comprehensive and accurate analysis of agricultural lands. Secondly, the incorporation of the Whale Optimization Algorithm enables the augmentation of collected data samples, leading to improved accuracy and reliability of the model. Thirdly, Deep CNNs allows the extraction of complex features from the data, leading to more accurate identification of agricultural lands. Finally, the use of Q-Learning ensures that the model is continuously optimized to improve its efficiency and effectiveness. The need for this work arises from the limitations of existing methods in accurately identifying agricultural lands. Traditional methods based on manual surveys and visual interpretation are time-consuming, expensive, and prone to errors. Moreover, existing automated methods often lack the ability to analyse multimodal satellite-based data samples and fail to provide accurate results. Based on these observations, the proposed augmented model offers a promising solution for identifying agricultural lands for sustainable development. The use of multimodal satellite-based data samples, WOA, Deep CNNs, and Q-Learning allows for an accurate and efficient analysis of agricultural lands, which can aid in sustainable development planning and decision-making operations. [ABSTRACT FROM AUTHOR]

Collaborative robots (cobots) are increasingly integrated into Industry 4.0 dynamic manufacturing environments that require frequent system reconfiguration due to changes in cobot paths and payloads. This necessitates fast methods for identifying payload inertial parameters to compensate the cobot controller and ensure precise and safe operation. Our prior work used Incremental Ensemble Model (IEM) to identify payload parameters, eliminating the need for an excitation path and thus removing the separate identification step. However, this approach suffers from catastrophic forgetting. This paper introduces a novel incremental ensemble learning method that addresses the problem of catastrophic forgetting by adding a new weak learner to the ensemble model for each new training bag. Moreover, it proposes a new classification model that assists the ensemble model in identifying which weak learner provides the most accurate estimation for new input data. The proposed method incrementally updates the identification model while the cobot navigates any task path, maintaining accuracy on old weak learner even after updating with new data. Validation performed on the Franka Emika cobot showcases the model's superior accuracy and adaptability, effectively eliminating the problem of catastrophic forgetting. [ABSTRACT FROM AUTHOR]

This paper explores the comparative analysis of federated learning (FL) and centralized learning (CL) models in the context of multi-class traffic flow classification for network applications, a timely study in the context of increasing privacy preservation concerns. Unlike existing literature that often omits detailed class-wise performance evaluation, and consistent data handling and feature selection approaches, our study rectifies these gaps by implementing a feed-forward neural network and assessing FL performance under both independent and identically distributed (IID) and non-independent and identically distributed (non-IID) conditions, with a particular focus on incremental training. In our cross-silo experimental setup involving five clients per round, FL models exhibit notable adaptability. Under IID conditions, the accuracy of the FL model peaked at 96.65%, demonstrating its robustness. Moreover, despite the challenges presented by non-IID environments, our FL models demonstrated significant resilience, adapting incrementally over rounds to optimize performance; in most scenarios, our FL models performed comparably to the idealistic CL model regarding multiple well-established metrics. Through a comprehensive traffic flow classification use case, this work (i) contributes to a better understanding of the capabilities and limitations of FL, offering valuable insights for the real-world deployment of FL, and (ii) provides a novel, large, carefully curated traffic flow dataset for the research community. [ABSTRACT FROM AUTHOR]

Just-in-time software defect prediction technology enables just-in-time defect prediction at the granularity of code changes, which is crucial for improving software code quality and ensuring software reliability. Traditional static software defect prediction models suffer from 'knowledge forgetting' when processing just-in-time software data streams, leading to poor model generalization performance. To address this, this paper proposed an incremental model method based on knowledge replay for just-in-time software defect prediction. Firstly, it used the knowledge replay mechanism stores model parameters and random samples to facilitate the learning of old knowledge. Secondly, this paper used a distributed training framework to perform incremental learning on just-in-time software data streams on local devices, achieving real-time model updates through restructuring. Lastly, this paper employed the knowledge distillation technique to construct a global incremental prediction model. Experiments show that this model performs better in terms of comprehensive performance compared to common modeling algorithms while ensuring training efficiency. [ABSTRACT FROM AUTHOR]

Functional linear regression is one of the main modeling tools for working with functional data. Since functional data are usually stream data essentially and there are some noises in functional data. Many numerical research studies of machine learning indicate that the noise samples not only increase the amount of storage space, but also affect the performance of algorithm. Therefore, in this paper we consider a new learning strategy by introducing incremental learning, Markov sampling for functional linear regression and propose a novel functional incremental linear square regression algorithm based on Markov sampling (FILSR-MS). To have a better understanding of the proposed FILSR-MS, we not only estimate the generalization bound of the proposed algorithm and establish the fast learning rate, but also present some useful discussions. The performance of the proposed algorithm is validated by the numerical experiments for benchmark repository. [ABSTRACT FROM AUTHOR]

In this paper, we improve the broad learning system (BLS) by speeding up the incremental learning for added inputs. We propose an efficient implementation for a step that is in the pseudoinverse computation of a partitioned matrix, to reduce the computational complexity. The proposed efficient implementation has two different forms for the cases of q > k and q ≤ k , respectively, where q and k denote the number of additional training samples and the total number of nodes, respectively. The proposed implementation for q > k utilizes the inverse of a sum of matrices to compute only a k × k matrix inverse, instead of a q × q matrix inverse in the original implementation, and the corresponding speedup for the matrix inversion operation in the number of floating-point operations is 1 2 (q / k) 3 . Moreover, it also speeds up two relevant matrix multiplication operations in the original implementation. On the other hand, the proposed implementation for q ≤ k speeds up one matrix multiplication operation in the original implementation. The numerical simulations show that both the proposed and original implementations always achieve the same testing accuracy. On the Modified National Institute of Standards and Technology dataset, the speedups of the proposed efficient BLS implementation over the original BLS implementation in total training time are 1.35–1.57 when q > k and 1.09–1.13 when q < k , while on the NYU object recognition benchmark dataset, the speedups are 1.10–1.35 when q > k and 1.07–1.09 when q < k. [ABSTRACT FROM AUTHOR]

Incremental learning is one of the effective methods of learning from the accumulated training samples and the large-scale dataset. The main advantages of incremental learning consist of making full use of historical information, reducing the training scale greatly and saving space and time consumption. Despite extensive research on incremental support vector machine (SVM) learning algorithms, most of them are based on independent and identically distributed samples (i.i.d.). Not only that, there has been no theoretical analysis of incremental SVM learning algorithms. In this paper, we mainly study the generalization bounds of this incremental SVM learning algorithm whose samples are based on uniformly geometric Markov chains, and exponentially strongly mixing sequence. As a special case, we also obtain the generalization bounds of i.i.d. samples. [ABSTRACT FROM AUTHOR]

Few-Shot Class Incremental Learning (FSCIL) is a trending topic in deep learning, addressing the need for models to incrementally learn novel classes, particularly in real-world scenarios where continuously emerging classes come with limited labeled samples. However, the majority of FSCIL research has been dedicated to image classification and object recognition tasks, with limited attention given to video action classification. In this paper, we present a new Cluster Compression and Generative Separation (CCGS) method for Incremental Few-Shot Video Action Recognition (iFSVAR), which introduces contrastive learning to boost the degree of class separation in the base session. Simultaneously, it creates numerous fine-grained classes with diverse semantics, effectively filling the unallocated representation space. Experimental results on UCF101, Kinetics, and Something-Something-V2 demonstrate the effectiveness of the framework. [ABSTRACT FROM AUTHOR]

This study introduces a two-step reinforcement learning (RL) strategy tailored for “The Lord of the Rings: The Card Game”, a complex multistage strategy card game. The research diverges from conventional RL methods by adopting a phased learning approach, beginning with a foundational learning step in a simplified version of the game and subsequently progressing to the complete, intricate game environment. This methodology notably enhances the AI agent’s adaptability and performance in the face of the unpredictable and challenging nature of the game. The paper also explores a multi-phase system where distinct RL agents are employed for various decision-making phases of the game. This approach has demonstrated remarkable improvement, with the RL agents achieving a winrate of 78.5% at the highest difficulty level. [ABSTRACT FROM AUTHOR]

Natural-language dialog is key for an intuitive human-robot interaction. It can be used not only to express humans' intents but also to communicate instructions for improvement if a robot does not understand a command correctly. It is of great importance to let robots learn from such interaction experiences in an incremental way to allow them to improve their behaviors or avoid mistakes in the future. In this paper, we propose a system to achieve such incremental learning of complex high-level behavior from natural interaction and demonstrate its implementation on a humanoid robot. Our system deploys large language models (LLMs) for high-level orchestration of the robot's behavior based on the idea of enabling the LLM to generate Python statements in an interactive console to invoke both robot perception and action. Human instructions, environment observations, and execution results are fed back to the LLM, thus informing the generation of the next statement. Since an LLM can misunderstand (potentially ambiguous) user instructions, we introduce incremental learning from the interaction, which enables the system to learn from its mistakes. For that purpose, the LLM can call another LLM responsible for code-level improvements in the current interaction based on human feedback. Subsequently, we store the improved interaction in the robot's memory so that it can later be retrieved on semantically similar requests. We integrate the system in the robot cognitive architecture of the humanoid robot ARMAR-6 and evaluate our methods both quantitatively (in simulation) and qualitatively (in simulation and real-world) by demonstrating generalized incrementally learned knowledge. [ABSTRACT FROM AUTHOR]

The outbreak of emerging infectious diseases poses significant challenges to global public health. Accurate early forecasting is crucial for effective resource allocation and emergency response planning. This study aims to develop a comprehensive predictive model for emerging infectious diseases, integrating the blending framework, transfer learning, incremental learning, and the biological feature Rt to increase prediction accuracy and practicality. By transferring features from a COVID-19 dataset to a monkeypox dataset and introducing dynamically updated incremental learning techniques, the model's predictive capability in data-scarce scenarios was significantly improved. The research findings demonstrate that the blending framework performs exceptionally well in short-term (7-day) predictions. Furthermore, the combination of transfer learning and incremental learning techniques significantly enhanced the adaptability and precision, with a 91.41% improvement in the RMSE and an 89.13% improvement in the MAE. In particular, the inclusion of the Rt feature enabled the model to more accurately reflect the dynamics of disease spread, further improving the RMSE by 1.91% and the MAE by 2.17%. This study underscores the significant application potential of multimodel fusion and real-time data updates in infectious disease prediction, offering new theoretical perspectives and technical support. This research not only enriches the theoretical foundation of infectious disease prediction models but also provides reliable technical support for public health emergency responses. Future research should continue to explore integrating data from multiple sources and enhancing model generalization capabilities to further enhance the practicality and reliability of predictive tools. [ABSTRACT FROM AUTHOR]

Global Positioning System (GPS)/Inertial Navigation System (INS) integration is a widely used technique for navigation and positioning applications. It combines the advantages of GPS and INS to provide accurate and reliable information. However, the GPS/INS integration suffers from performance degradation during a GPS outage, which occurs when natural or artificial factors block the GPS signal. The novelty of this paper is improving GPS/INS integration performance during GPS outages using Incremental Regularized LSTM (IncRLSTM) learning. Incremental learning is a learning paradigm that can be learned from streaming data online and updating the model parameters without forgetting the previous ones. Also, regularization is a technique that prevents overfitting and improves the generalization of the network by adding some constraints or penalties to the model. IncRLSTM learning models the GPS signal as a multi-objective regression and corrects the INS output with the Faded Memory Kalman filter. The results on real-world datasets significantly show the reduction of positioning errors by an average of 72 % during GPS outages and the improvement of the accuracy and robustness of GPS/INS integration by an average of 58 % compared with existing methods. Moreover, IncRLSTM presents, on average, a 10 % improvement compared to the existing methods. [ABSTRACT FROM AUTHOR]

Electroslag remelting (ESR) is essential for producing high‐end special steel, but its complex process and numerous influencing factors make quality control challenging. This study addresses oxygen content control during ESR using a big data machine learning approach. An incremental learning strategy is proposed based on an optimized Wasserstein generative adversarial network with gradient penalty (WGAN‐GP) for data enhancement, focusing on G20Cr2Ni4A bearing steel. The WGAN‐GP model enhances time‐series data and metadata, utilizing long short‐term memory networks, fully connected networks, and attention mechanisms. The effectiveness of data enhancement is verified using a deep neural network classifier and statistical methods. Data is divided into historical and data streams, with an incremental learning strategy based on histogram gradient boosting regression trees to prevent catastrophic forgetting and improve efficiency through knowledge distillation and real‐time hyperparameter adjustment. Results show that the data augmentation method significantly improves model generalization and accuracy in small sample metallurgy. The incremental learning strategy enhances prediction accuracy for oxygen content, contributing to better cleanliness quality of electroslag steel. This study offers a novel approach for addressing small sample challenges in metallurgical processes. [ABSTRACT FROM AUTHOR]

Modern natural language processing (NLP) state-of-the-art (SoTA) deep learning (DL) models have hundreds of millions of parameters, making them extremely complex. Large datasets are required for training these models, and while pretraining has reduced this requirement, human-labelled datasets are still necessary for fine-tuning. Few-shot learning (FSL) techniques, such as meta-learning, try to train models from smaller datasets to mitigate this cost. However, the tasks used to evaluate these meta-learners frequently diverge from the problems in the real world that they are meant to resolve. This work aims to apply meta-learning to a problem that is more pertinent to the real world: class incremental learning (IL). In this scenario, after completing its training, the model learns to classify newly introduced classes. One unique quality of meta-learners is that they can generalise from a small sample size to classes that have never been seen before, which makes them especially useful for class incremental learning (IL). The method describes how to emulate class IL using proxy new classes. This method allows a meta-learner to complete the task without the need for retraining. To generate predictions, the transformer-based aggregation function in a meta-learner that modifies data from examples across all classes has been proposed. The principal contributions of the model include concurrently considering the entire support and query sets, and prioritising attention to crucial samples, such as the question, to increase the significance of its impact during inference. The outcomes demonstrate that the model surpasses prevailing benchmarks in the industry. Notably, most meta-learners demonstrate significant generalisation in the context of class IL even without specific training for this task. This paper establishes a high-performing baseline for subsequent transformer-based aggregation techniques, thereby emphasising the practical significance of meta-learners in class IL. [ABSTRACT FROM AUTHOR]

The incremental recommendation involves updating existing models by extracting information from interaction data at current time-step, with the aim of maintaining model accuracy while addressing limitations including parameter dependencies and inefficient training. However, real-time user interaction data is often afflicted by substantial noise and invalid samples, presenting the following key challenges for incremental model updating: (1) how to effectively extract valuable new knowledge from interaction data at the current time-step to ensure model accuracy and timeliness, and (2) how to safeguard against the catastrophic forgetting of long-term stable preference information, thus preserving the model's sensitivity during cold-starts. In response to these challenges, we propose the Incremental Recommendation with Stable Latent Side-information Updating (SIIFR). This model employs a side-information augmenter to extract valuable latent side-information from user interaction behavior at time-step T, thereby sidestepping the interference caused by noisy interaction data and acquiring stable user preference. Moreover, the model utilizes rough interaction data at time-step T + 1 , in conjunction with existing side-information enhancements to achieve incremental updates of latent preferences, thereby ensuring the model's efficacy during cold-start. Furthermore, SIIFR leverages the change rate in user latent side-information to mitigate catastrophic forgetting that results in the loss of long-term stable preference information. The effectiveness of the proposed model is validated and compared against existing models using four popular incremental datasets. The model code can be achieved at: https://github.com/LNNU-computer-research-526/FR-sii. [ABSTRACT FROM AUTHOR]

Modulation recognition is an important technology used to correctly identify the modulation modes of wireless signals and is widely used in cooperative and confrontational scenarios. Traditional modulation-recognition algorithms require the assistance of expert experiences, which constrains their applications. With the rapid development of artificial intelligence in recent years, deep learning (DL) is widely advocated for intelligent modulation recognition. Typically, DL-based modulation-recognition algorithms implicitly assume a relatively static scenario in which the signal samples of all the modulation modes can be completely collected in advance. In practical situations, the radio environment is quite dynamic and the signal samples with new modulation modes may appear sequentially, in which the current DL-based modulation-recognition algorithms may require unacceptable time and computing resource consumption to re-train the optimal model from scratch. In this study, we leveraged incremental learning (IL) and designed a novel IL-based modulation-recognition algorithm that consists of an initial stage and multiple incremental stages. The main novelty of the proposed algorithm lies in the new loss function design in each incremental stage, which combines the distillation loss of recognizing old modulation modes and the cross-entropy loss of recognizing new modulation modes. With the proposed algorithm, the knowledge of the signal samples with new modulation modes can be efficiently learned in the current stage without forgetting the knowledge learned in the previous stages. The simulation results demonstrate that the proposed algorithm could achieve a recognition accuracy close to the upper bound with a much lower computing time and it outperformed the existing IL-based benchmarks. [ABSTRACT FROM AUTHOR]

In recent years, semi-supervised learning (SSL) methods based on pseudo-labeling algorithms have been widely applied and achieved significant success. However, most existing deep semi-supervised learning methods suffer from the problem of distribution shift between the source and target domains, as well as the issue of "cognitive bias" in pseudo labeling algorithms, where the model's errors are difficult to rectify as they accumulate through the pseudo-labeling process. This paper introduces the concept of causal invariance and proposes an incremental repeated labeling strategy with a high confidence threshold to enhance the utilization of unlabeled samples. It effectively solves the issue of distribution discrepancy between the source and target domains in the field of semi-supervised learning, as well as the problem of pseudo label "cognitive bias", thus improving the accuracy of the model. Extensive experiments on CIFAR-10, CIFAR-100, SVHN, STL-10, PACS and VLCS demonstrate that semi-supervised incremental models based on causal invariance have a significant improvement in domain generalization ability compared with state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Building systems for predicting human socio-emotional states has promising applications; however, if trained on biased data, such systems could inadvertently yield biased decisions. Bias mitigation remains an open problem, which tackles the correction of a model's disparate performance over different groups defined by particular sensitive attributes (e.g., gender, age, and race). In this work, we design a novel fairness loss function named Multi-Group Parity (MGP) to provide a generalised approach for bias mitigation in personality computing. In contrast to existing works in the literature, MGP is generalised as it features four ‘multiple’ properties (4Mul): multiple tasks, multiple modalities, multiple sensitive attributes, and multi-valued attributes. Moreover, we explore how to incrementally mitigate the biases when more sensitive attributes are taken into consideration sequentially. Towards this problem, we introduce a novel algorithm that utilises an incremental learning framework to mitigate bias against one attribute data at a time without compromising past fairness. Extensive experiments on two large-scale multi-modal personality recognition datasets validate the effectiveness of our approach in achieving superior bias mitigation under the proposed four properties and incremental debiasing settings. [ABSTRACT FROM AUTHOR]

A modified fuzzy min-max classifier is presented that differs from the original in the way that the hyperbox expansion operation is performed. The classifier has been tested on the solution of cybersecurity problems, such as detecting spam, phishing sites and attacks on network connections. The results of experiments results showed an improvement in the accuracy relative to the original fuzzy min-max classifier. Comparisons with six alternative incremental learning classifiers showed competitive results on the false acceptance rate, the false reject rate, and the F1-score values. [ABSTRACT FROM AUTHOR]

Event detection is a crucial task in information extraction. Existing research primarily focuses on machine automatic detection tasks, which often perform poorly in certain practical applications. To address this, an interactive event-detection mode of "machine recommendation-human review–machine incremental learning" was proposed. In this mode, we study a few-shot continual class-incremental learning scenario, where the challenge is to learn new-class events with limited samples while preserving memory of old class events. To tackle these challenges, we propose a class-incremental learning method for interactive event detection via Interaction, Contrast and Distillation (ICD). We design a replay strategy based on representative and confusable samples to retain the most valuable samples under limited conditions; we introduce semantic-boundary-smoothness contrastive learning for effective learning of new-class events with few samples; and we employ hierarchical distillation to mitigate catastrophic forgetting. These methods complement each other and show strong performance. Experimental results demonstrate that, in the 5-shot 5-round class incremental-learning settings on two Chinese event-detection datasets ACE and DuEE, our method achieves final recall rates of 71.48% and 90.39%, respectively, improving by 6.86% and 3.90% over the best baseline methods. [ABSTRACT FROM AUTHOR]