Continuous Learning Framework Research Articles

A core space gradient projection-based continual learning framework for remaining useful life prediction of machinery under variable operating conditions

Recently, continual learning has received particular attention in machinery remaining useful life (RUL) prediction, which enables prognostics networks to gradually improve performance without laborious retraining. Existing studies, however, have the following limitations: 1) The impacts of variable operating conditions are not explicitly considered during continual learning. 2) The continual learnability of prognostics network is limited by the lack of knowledge compression and de-redundancy. To overcome the abovementioned limitations, a novel continual learning framework is proposed for RUL prediction of machinery under variable operating conditions. First, a multi-kernel swarm convolution block is devised to automatically capture degradation features during continual learning without increasing computational consumption, which combines convolutional operations at various scales and focuses attention on appropriate scales based on operating condition information. Then, core space gradient projection is proposed for continual learning of prognostics networks, which mitigates forgetting by guiding gradient descent along the orthogonal direction of the previously input subspace. This approach also ensures the network learnability during continual learning by knowledge compression and de-redundancy to identify core space. The proposed framework is verified using accelerated degradation datasets of rolling element bearings with variable operating conditions. Experimental results show that the proposed framework is superior to some existing continual learning-based prognostics approaches for RUL prediction under variable operating conditions.

DOCTOR: A Multi-Disease Detection Continual Learning Framework Based on Wearable Medical Sensors

DOCTOR: A Multi-Disease Detection Continual Learning Framework Based on Wearable Medical Sensors

Analysis of Continual Learning Techniques for Image Generative Models with Learned Class Information Management.

The advancements in deep learning have significantly enhanced the capability of image generation models to produce images aligned with human intentions. However, training and adapting these models to new data and tasks remain challenging because of their complexity and the risk of catastrophic forgetting. This study proposes a method for addressing these challenges involving the application of class-replacement techniques within a continual learning framework. This method utilizes selective amnesia (SA) to efficiently replace existing classes with new ones while retaining crucial information. This approach improves the model's adaptability to evolving data environments while preventing the loss of past information. We conducted a detailed evaluation of class-replacement techniques, examining their impact on the "class incremental learning" performance of models and exploring their applicability in various scenarios. The experimental results demonstrated that our proposed method could enhance the learning efficiency and long-term performance of image generation models. This study broadens the application scope of image generation technology and supports the continual improvement and adaptability of corresponding models.

Recall-Oriented Continual Learning with Generative Adversarial Meta-Model

The stability-plasticity dilemma is a major challenge in continual learning, as it involves balancing the conflicting objectives of maintaining performance on previous tasks while learning new tasks. In this paper, we propose the recalloriented continual learning framework to address this challenge. Inspired by the human brain’s ability to separate the mechanisms responsible for stability and plasticity, our framework consists of a two-level architecture where an inference network effectively acquires new knowledge and a generative network recalls past knowledge when necessary. In particular, to maximize the stability of past knowledge, we investigate the complexity of knowledge depending on different representations, and thereby introducing generative adversarial meta-model (GAMM) that incrementally learns task-specific parameters instead of input data samples of the task. Through our experiments, we show that our framework not only effectively learns new knowledge without any disruption but also achieves high stability of previous knowledge in both task-aware and task-agnostic learning scenarios. Our code is available at: https://github.com/bigdata-inha/recall-orientedcl-framework.

Continual learning framework for a multicenter study with an application to electrocardiogram

Deep learning has been increasingly utilized in the medical field and achieved many goals. Since the size of data dominates the performance of deep learning, several medical institutions are conducting joint research to obtain as much data as possible. However, sharing data is usually prohibited owing to the risk of privacy invasion. Federated learning is a reasonable idea to train distributed multicenter data without direct access; however, a central server to merge and distribute models is needed, which is expensive and hardly approved due to various legal regulations. This paper proposes a continual learning framework for a multicenter study, which does not require a central server and can prevent catastrophic forgetting of previously trained knowledge. The proposed framework contains the continual learning method selection process, assuming that a single method is not omnipotent for all involved datasets in a real-world setting and that there could be a proper method to be selected for specific data. We utilized the fake data based on a generative adversarial network to evaluate methods prospectively, not ex post facto. We used four independent electrocardiogram datasets for a multicenter study and trained the arrhythmia detection model. Our proposed framework was evaluated against supervised and federated learning methods, as well as finetuning approaches that do not include any regulation to preserve previous knowledge. Even without a central server and access to the past data, our framework achieved stable performance (AUROC 0.897) across all involved datasets, achieving comparable performance to federated learning (AUROC 0.901).

Facing spatiotemporal heterogeneity: A unified federated continual learning framework with self-challenge rehearsal for industrial monitoring tasks

Federated learning (FL) has attracted significant interest in industrial monitoring tasks. Practically, FL-based methods face spatiotemporal heterogeneity, including the inter-client distribution discrepancy spatially and intra-client distribution change temporally. Some methods introduce continual learning into FL methods, which aims to facilitate the models to preserve historical knowledge. However, most of them ignore the interference between previous and collective knowledge caused by spatiotemporal heterogeneity, leading to inaccurate historical memories and unstable aggregation. This study proposes a unified federated continual learning framework for industrial monitoring tasks, which introduces an “intra-rehearsal and inter-aggregation” mechanism to address spatiotemporal heterogeneity. To recall historical distributions for continual learning capability, a self-challenge diffusion and replay model is proposed by designing masking recovery tasks with concurrent random scale and position. It is proven to reduce the conditional entropy and improve the ability to portray complex distributions. To reduce parameter variations during federated aggregation, a gradient balance policy is designed by applying iteration-related regularity to the cumulative gradient within rounds. It is theoretically analyzed to reduce the upper bound of the minimal global gradient. Our unified framework is applicable to supervised and unsupervised paradigms and thus covers typical industrial monitoring tasks. Experiments on a multiphase flow process evaluate the continual learning performance of the proposed framework. Compared to the sub-optimal algorithms, our ERR decreases by 9.20% and 10.60%, and FWT decreases by 0.039 and 0.030 in RMSE and MAE, respectively. In addition, the experiments are conducted to demonstrate the superiority of rehearsal ability and training stability.

Large-scale continual learning for ancient Chinese character recognition

Ancient Chinese character recognition is a challenging problem in the field of pattern recognition. It is difficult to collect all character classes during the training stage due to the numerous classes of ancient Chinese characters and the likelihood of discovering new characters over time. A solution to address this problem is continual learning. However, most continual learning methods are not well-suited for large-scale applications, making them insufficient for solving the problem of ancient Chinese character recognition. Although saving raw data for old classes is a good approach for continual learning to address large-scale problems, it is often infeasible due to the lack of data accessibility in reality. To solve these problems, we propose a large-scale continual learning framework based on the convolutional prototype network (CPN), which does not save raw data for old classes. In this paper, several basic strategies have been proposed for the initial training stage to enhance the feature extraction ability and robustness of the network, which can improve the performance of the model in continual learning. In addition, we propose two practical methods in varying feature space (parameters of feature extractor are changeable) and fixed feature space (parameters of feature extractor are fixed), which enable the model to carry out large-scale continual learning. The proposed method does not save the raw data of old classes and enables simultaneous classification of all existing classes without knowing the incremental batch number. Experiments on the CASIA-AHCDB dataset with 5000 character classes demonstrate the effectiveness and superiority of the proposed method.

A continuous learning approach to brain tumor segmentation: integrating multi-scale spatial distillation and pseudo-labeling strategies.

This study presents a novel continuous learning framework tailored for brain tumour segmentation, addressing a critical step in both diagnosis and treatment planning. This framework addresses common challenges in brain tumour segmentation, such as computational complexity, limited generalisability, and the extensive need for manual annotation. Our approach uniquely combines multi-scale spatial distillation with pseudo-labelling strategies, exploiting the coordinated capabilities of the ResNet18 and DeepLabV3+ network architectures. This integration enhances feature extraction and efficiently manages model size, promoting accurate and fast segmentation. To mitigate the problem of catastrophic forgetting during model training, our methodology incorporates a multi-scale spatial distillation scheme. This scheme is essential for maintaining model diversity and preserving knowledge from previous training phases. In addition, a confidence-based pseudo-labelling technique is employed, allowing the model to self-improve based on its predictions and ensuring a balanced treatment of data categories. The effectiveness of our framework has been evaluated on three publicly available datasets (BraTS2019, BraTS2020, BraTS2021) and one proprietary dataset (BraTS_FAHZU) using performance metrics such as Dice coefficient, sensitivity, specificity and Hausdorff95 distance. The results consistently show competitive performance against other state-of-the-art segmentation techniques, demonstrating improved accuracy and efficiency. This advance has significant implications for the field of medical image segmentation. Our code is freely available at https://github.com/smallboy-code/A-brain-tumor-segmentation-frameworkusing-continual-learning.

A Collective and Continuous Rule Learning Framework towards Enhanced Driving Safety and Decision Transparency

A Collective and Continuous Rule Learning Framework towards Enhanced Driving Safety and Decision Transparency

Online Active Continual Learning for Robotic Lifelong Object Recognition.

In real-world applications, robotic systems collect vast amounts of new data from ever-changing environments over time. They need to continually interact and learn new knowledge from the external world to adapt to the environment. Particularly, lifelong object recognition in an online and interactive manner is a crucial and fundamental capability for robotic systems. To meet this realistic demand, in this article, we propose an online active continual learning (OACL) framework for robotic lifelong object recognition, in the scenario of both classes and domains changing with dynamic environments. First, to reduce the labeling cost as much as possible while maximizing the performance, a new online active learning (OAL) strategy is designed by taking both the uncertainty and diversity of samples into account to protect the information volume and distribution of data. In addition, to prevent catastrophic forgetting and reduce memory costs, a novel online continual learning (OCL) algorithm is proposed based on the deep feature semantic augmentation and a new loss-based deep model and replay buffer update, which can mitigate the class imbalance between the old and new classes and alleviate confusion between two similar classes. Moreover, the mistake bound of the proposed method is analyzed in theory. OACL allows robots to select the most representative new samples to query labels and continually learn new objects and new variants of previously learned objects from a nonindependent and identically distributed (i.i.d.) data stream without catastrophic forgetting. Extensive experiments conducted on real lifelong robotic vision datasets demonstrate that our algorithm, even trained with fewer labeled samples and replay exemplars, can achieve state-of-the-art performance on OCL tasks.

Expanding continual few-shot learning benchmarks to include recognition of specific instances.

Continual learning and few-shot learning are important frontiers in progress toward broader Machine Learning (ML) capabilities. Recently, there has been intense interest in combining both. One of the first examples to do so was the Continual few-shot Learning (CFSL) framework of Antoniou et al. (2020). In this study, we extend CFSL in two ways that capture a broader range of challenges, important for intelligent agent behaviour in real-world conditions. First, we increased the number of classes by an order of magnitude, making the results more comparable to standard continual learning experiments. Second, we introduced an 'instance test' which requires recognition of specific instances of classes-a capability of animal cognition that is usually neglected in ML. For an initial exploration of ML model performance under these conditions, we selected representative baseline models from the original CFSL work and added a model variant with replay. As expected, learning more classes is more difficult than the original CFSL experiments, and interestingly, the way in which image instances and classes are presented affects classification performance. Surprisingly, accuracy in the baseline instance test is comparable to other classification tasks, but poor given significant occlusion and noise. The use of replay for consolidation substantially improves performance for both types of tasks, but particularly for the instance test.

Continual Learning, Fast and Slow.

According to the Complementary Learning Systems (CLS) theory (McClelland et al. 1995) in neuroscience, humans do effective continual learning through two complementary systems: a fast learning system centered on the hippocampus for rapid learning of the specifics, individual experiences; and a slow learning system located in the neocortex for the gradual acquisition of structured knowledge about the environment. Motivated by this theory, we propose DualNets (for Dual Networks), a general continual learning framework comprising a fast learning system for supervised learning of pattern-separated representation from specific tasks and a slow learning system for representation learning of task-agnostic general representation via Self-Supervised Learning (SSL). DualNets can seamlessly incorporate both representation types into a holistic framework to facilitate better continual learning in deep neural networks. Via extensive experiments, we demonstrate the promising results of DualNets on a wide range of continual learning protocols, ranging from the standard offline, task-aware setting to the challenging online, task-free scenario. Notably, on the CTrL (Veniat et al. 2020) benchmark that has unrelated tasks with vastly different visual images, DualNets can achieve competitive performance with existing state-of-the-art dynamic architecture strategies (Ostapenko et al. 2021). Furthermore, we conduct comprehensive ablation studies to validate DualNets efficacy, robustness, and scalability.

Adaptive instance similarity embedding for online continual learning

We study the online continual learning (CL) paradigm, where the learner must continually learn a sequence of tasks. In this setting, improving the learning ability of the model and mitigating catastrophic forgetting are two pivotal factors. Note that most existing approaches for online continual learning are based on the experience replay strategy. In this type of method, a memory buffer is applied to store a subset of previous tasks to prevent catastrophic forgetting. The samples from between the current task and the memory buffer are jointly trained to update the network parameters. Consider that most methods only generate the feature embeddings via a shared feature extractor and then train the network via cross-entropy loss. We argue that such methods fail to explore the feature embedding in its entirety and neglect the similar relations between samples, thus leading to lower discriminant performance, especially in an online learning setting. To this end, we propose the Adaptive Instance Similarity Embedding for Online Continual Learning (AISEOCL) framework, which further takes all the sample relations in a given batch into account. In detail, firstly, the experience replay strategy is used to avoid catastrophic forgetting. Then, during training, we apply the adaptive similar embedding to obtain additional valuable similar information from the current training samples composed of the current and previous tasks. Since not all samples are equally important to make a prediction, we further weigh the importance of each instance accordingly resorting to the attention mechanism. Importantly, we further impose a similarity distillation loss on the distributions of the similarity relationship between current and previous models. Such operation can transfer the similarity relationship between different samples from the old model to the current model to alleviate catastrophic forgetting. With this strategy, AISEOCL can further improve the learning ability of the model while enhancing the discriminant power, which is also beneficial to stably resist forgetting. The experiments on several existing benchmarks validate the effectiveness of our proposed approach.

FedViT: Federated continual learning of vision transformer at edge

Deep Neural Networks (DNNs) have been ubiquitously adopted in internet of things and are becoming an integral part of our daily life. When tackling the evolving learning tasks in real world, such as classifying different types of objects, DNNs face the challenge to continually retrain themselves according to the tasks on different edge devices. Federated continual learning (FCL) is a promising technique that offers partial solutions but yet to overcome the following difficulties: the significant accuracy loss due to the limited on-device processing, the negative knowledge transfer caused by the limited communication of non-IID (non-Independent and Identically Distributed) data, and the limited scalability on the tasks and edge devices. Moreover, existing FCL techniques are designed for convolutional neural networks (CNNs), which have not utilized the full potential of newly emerged powerful vision transformers (ViTs). Considering ViTs depend heavily on training data diversity and volume, we hypothesize ViTs are well-suited for FCL where data arrives continually. In this paper, we propose FedViT, an accurate and scalable federated continual learning framework for ViT models, via a novel concept of signature task knowledge. FedViT is a client-side solution that continuously extracts and integrates the knowledge of signature tasks which are highly influenced by the current task. Each client of FedViT is composed of a knowledge extractor, a gradient restorer and, most importantly, a gradient integrator. Upon training for a new task, the gradient integrator ensures the prevention of catastrophic forgetting and mitigation of negative knowledge transfer by effectively combining signature tasks identified from the past local tasks and other clients’ current tasks through the global model. We implement FedViT in PyTorch and extensively evaluate it against state-of-the-art techniques using popular federated continual learning benchmarks. Extensive evaluation results on heterogeneous edge devices show that FedViT improves model accuracy by 88.61% without increasing model training time, reduces communication cost by 61.55%, and achieves more improvements under difficult scenarios such as large numbers of tasks or clients, and training different complex ViT models.

Personalized Federated Continual Learning for Task-Incremental Biometrics

In the age of Internet of Things where information is explosively growing, people pay more attention on personal privacy. In the real-world task-incremental scenario for biometrics, every edge device faces continuous task flows of private data without communication with others. Security and Performance are the primary concerns in identity authentication, and Federated Continual Learning (FCL) is a promising solution. In this paper, we design a personalized federated continual learning framework to solve the problem of sequentially identification in every distributed device. For each client, we create an adaptive continual meta-learning model called cTD-αMAML, aiming to align the gradients of previous and new tasks and to make the learning-rate model learnable. For central aggregation, the server gathers the meta-initialization from every local update and allocates the updated global meta-initialization to clients. We propose an extension of FedAvg to locally reserve the learnable learning-rate network to realize the personalization of clients. Results prove that in Continual Learning, our cTD-αMAML can learn to learn the seen tasks and avoid catastrophic forgetting. And in Federated Continual Learning, our personalized method realizes the knowledge transferring across clients, meanwhile improving the local performance and reducing the communication cost. In this way, the proposed personalized federated continual learning framework can obtain a biometric template that is able to learn the expression space for new tasks with rapid adaption.

Federated Continuous Learning Based on Stacked Broad Learning System Assisted by Digital Twin Networks: An Incremental Learning Approach for Intrusion Detection in UAV Networks

The edge of the Internet of Things (IoT), which consists of Unmanned Aerial Vehicles (UAVs), is vulnerable to network intrusion because software and wireless connections are used extensively in the IoT. Designing an efficient Intrusion Detection System (IDS) model is imperative. However, when creating IDS models with distributed data collected by UAVs, it is necessary to take precautions to protect the data’s security and privacy. Furthermore, most of the IDS models are focused on one-time learning but not on continuous learning. To this end, we propose a Federated Continuous learning framework with a Stacked Broad Learning System (FCL-SBLS) based on Digital Twin Network (DTN), which can learn and train the IDS model on new data quickly and continuously. In order to improve the efficiency and quality of the IDS model when training and aggregation, we employ an asynchronous federated learning architecture, and a Deep Deterministic Policy Gradient (DDPG)-based UAV selection scheme assisted by DTN is proposed to help the global IDS model aggregation. The presented algorithm is validated using the CIC-IDS2017 dataset, and the simulation results reveal that our algorithm achieves higher efficiency and accuracy than the existing FL scheme.

Self-driven continual learning for class-added motor fault diagnosis based on unseen fault detector and propensity distillation

Continual learning is a high-potential technique that enables intelligent motor fault diagnosis models to extend new diagnosable fault classes without costly training from scratch. However, existing continual learning methods have the following limitations. (1) They manually detect new faults, which is labor-intensive, untimely, and more importantly, may lead to mistaken diagnosis results. (2) They adopt the traditional knowledge distillation to align the absolute responses of old and new models, which alleviates catastrophic forgetting but restricts flexible learning from incremental datasets. To overcome the above limitations, this paper proposes a novel self-driven continual learning framework for class-added motor fault diagnosis, which can spontaneously detect unseen faults and perform more flexible continual learning from incremental datasets. For the automatic detection of unseen faults, after collecting online samples, adversarial training with exemplars of each seen class is conducted to measure the class separability. The truth fault classes that are unseen for diagnosis models can be clearly distinguished from all seen classes, and correspondingly missed diagnosis or misdiagnosing can be avoided effectively and incremental samples with new fault types can be collected quickly. For the flexible continual learning strategy, a more flexible knowledge distillation is proposed to preserve the prediction propensity rather than the absolute response. This strategy not only keeps the recognition performance of old classes but also loosens unnecessary constraints and increases the diagnosis model plasticity to learn new knowledge from incremental datasets, thus improving the accuracy of motor fault diagnosis during continual learning. The effectiveness of the proposed method is verified by conducting fault simulation experiments of three-phase motors and its superiority is also demonstrated by comparing it with some state-of-the-art diagnosis methods.

Continual learning approaches for single cell RNA sequencing data

Single-cell RNA sequencing data is among the most interesting and impactful data of today and the sizes of the available datasets are increasing drastically. There is a substantial need for learning from large datasets, causing nontrivial challenges, especially in hardware. Loading even a single dataset into the memory of an ordinary, off-the-shelf computer can be infeasible, and using computing servers might not always be an option. This paper presents continual learning as a solution to such hardware bottlenecks. The findings of cell-type classification demonstrate that XGBoost and Catboost algorithms, when implemented in a continual learning framework, exhibit superior performance compared to the best-performing static classifier. We achieved up to 10% higher median F1 scores than the state-of-the-art on the most challenging datasets. On the other hand, these algorithms can suffer from variations in data characteristics across diverse datasets, pointing out indications of the catastrophic forgetting problem.

CRNet: A Fast Continual Learning Framework With Random Theory.

Artificial neural networks are prone to suffer from catastrophic forgetting. Networks trained on something new tend to rapidly forget what was learned previously, a common phenomenon within connectionist models. In this work, we propose an effective and efficient continual learning framework using random theory, together with Bayes' rule, to equip a single model with the ability to learn streaming data. The core idea of our framework is to preserve the performance of old tasks by guiding output weights to stay in a region of low error while encountering new tasks. In contrast to the existing continual learning approaches, our main contributions concern (1) closed-formed solutions with detailed theoretical analysis; (2) training continual learners by one-pass observation of samples; (3) remarkable advantages in terms of easy implementation, efficient parameters, fast convergence, and strong task-order robustness. Comprehensive experiments under popular image classification benchmarks, FashionMNIST, CIFAR-100, and ImageNet, demonstrate that our methods predominately outperform the extensive state-of-the-art methods on training speed while maintaining superior accuracy and the number of parameters, in the class incremental learning scenario. Code is available at https://github.com/toil2sweet/CRNet.

Continual Learning for Cuffless Blood Pressure Measurement using PPG and ECG Signals.

Although numerous studies have been conducted on cuffless blood pressure (BP) estimation using machine learning methods, most of the data-driven models are static, with model parameters fixed after training is complete. However, BP is dynamic and the performance would degrade for a static model when the to-be predicted BP distribution deviates from the training BP distribution. In this paper, we propose a continual learning (CL) framework in which deep learning models are developed to learn dynamically and continuously for arterial BP (ABP) estimation with photoplethysmography (PPG) and electrocardiogram (ECG) waveforms. The effectiveness of the CL model is validated on UCI Repository and MIMIC-III database with a total of 132 individual samples, and compared with conventional training method. It was found that the CL model improved the ABP estimation accuracy in terms of mean absolute error (MAE) by 17.47% on average compared with conventional training model. Furthermore, the improvement increased with the variability of ABP. These results demonstrate that CL model has potential to estimate dynamic ABP, which has been challenging with conventional training.