Quantum continual learning aims to train quantum models on sequential tasks without losing previously learned knowledge. However, variational quantum classifiers (VQCs) are prone to catastrophic forgetting under nonstationary task distributions. We propose quantum elastic weight consolidation (QEWC), a quantum Fisher information (QFI)-informed regularization method for mitigating forgetting. Unlike conventional elastic weight consolidation based on classical Fisher information (CFI), which measures parameter importance through measurement-dependent output statistics, QEWC uses QFI to quantify the intrinsic sensitivity of the parameterized quantum state. This gives an information-geometric view in which important parameters are identified by the local response of the quantum state manifold. We evaluate QEWC on VQCs trained on sequential binary classification tasks, including classical image-classification and quantum phase-classification tasks. Simulations show that sequential training without regularization causes severe forgetting, while both CFI-based EWC and QFI-based QEWC improve retention of previous tasks. Mechanistic analyses further show that the two methods impose different regularization geometries: CFI acts selectively on measurement-sensitive directions, whereas QFI imposes a denser state-geometric constraint over parameter space. Under depolarizing noise, CFI values are strongly suppressed by degraded measurement statistics, while QFI preserves a more stable sensitivity structure of the noisy parameterized quantum state. These results establish QEWC as a physically motivated approach for studying and mitigating forgetting in quantum continual learning through quantum-state geometry.
Quantum continual learning aims to train quantum models on sequential tasks without losing previously learned knowledge. However, variational quantum classifiers (VQCs) are prone to catastrophic forgetting under nonstationary task distributions. We propose quantum elastic weight consolidation (QEWC), a quantum Fisher information (QFI)-informed regularization method for mitigating forgetting. Unlike conventional elastic weight consolidation based on classical Fisher information (CFI), which measures parameter importance through measurement-dependent output statistics, QEWC uses QFI to quantify the intrinsic sensitivity of the parameterized quantum state. This gives an information-geometric view in which important parameters are identified by the local response of the quantum state manifold. We evaluate QEWC on VQCs trained on sequential binary classification tasks, including classical image-classification and quantum phase-classification tasks. Simulations show that sequential training without regularization causes severe forgetting, while both CFI-based EWC and QFI-based QEWC improve retention of previous tasks. Mechanistic analyses further show that the two methods impose different regularization geometries: CFI acts selectively on measurement-sensitive directions, whereas QFI imposes a denser state-geometric constraint over parameter space. Under depolarizing noise, CFI values are strongly suppressed by degraded measurement statistics, while QFI preserves a more stable sensitivity structure of the noisy parameterized quantum state. These results establish QEWC as a physically motivated approach for studying and mitigating forgetting in quantum continual learning through quantum-state geometry.
Quantum computers may outperform classical computers on machine learning tasks. Yet, quantum learning systems may suffer from catastrophic forgetting, which is widely believed to be an obstacle to achieving continual learning. Here, we report an experimental demonstration of quantum continual learning on a superconducting processor. In particular, we sequentially train a quantum classifier with three tasks, two about identifying real-life images and one on classifying quantum states, and demonstrate its catastrophic forgetting. To overcome this dilemma, we exploit the elastic weight consolidation strategy and show that the quantum classifier can incrementally retain knowledge across three tasks with an average accuracy exceeding 92.3%. Additionally, for sequential tasks involving quantum-engineered data, we demonstrate that the quantum classifier outperforms a classical classifier with a comparable number of parameters. Our results establish a viable strategy for empowering quantum learning systems with adaptability to sequential tasks.
We introduce CL-QAS, a continual quantum architecture search framework that mitigates the challenges of costly amplitude encoding and catastrophic forgetting in variational quantum circuits. The method uses Tensor-Train encoding to efficiently compress high-dimensional stochastic signals into low-rank quantum feature representations. A bi-loop learning strategy separates circuit parameter optimization from architecture exploration, while an Elastic Weight Consolidation regularization ensures stability across sequential tasks. We derive theoretical upper bounds on approximation, generalization, and robustness under quantum noise, demonstrating that CL-QAS achieves controllable expressivity, sample-efficient generalization, and smooth convergence without barren plateaus. Empirical evaluations on electrocardiogram (ECG)-based signal classification and financial time-series forecasting confirm substantial improvements in accuracy, balanced accuracy, F1 score, and reward. CL-QAS maintains strong forward and backward transfer and exhibits bounded degradation under depolarizing and readout noise, highlighting its potential for adaptive, noise-resilient quantum learning on near-term devices.
Catastrophic forgetting emerges when a neural network’s parameters undergo continuous updates during the sequential training of multiple tasks. The ongoing adaptation, while enhancing the model’s suitability for new tasks, inadvertently leads to a degradation in performance on previously learned tasks. This challenge significantly impedes the sequential learning capabilities essential for the advancement of artificial general intelligence. The phenomenon of catastrophic forgetting also occurs in the field of quantum machine learning, where parametric quantum circuits serve a role analogous to neural networks. In this study, our focus is to explore strategies for mitigating catastrophic forgetting within quantum learning models for quantum data. In this context, we employ a task-based hard attention mechanism, which automatically generates masks for each task to regulate the network’s learning process and resist catastrophic forgetting. To our knowledge, the proposed method is the first one in the field of quantum machine learning to adjust the model structure to prevent catastrophic forgetting. Numerical simulation results demonstrate that our approach preserves the model’s performance on previous tasks without compromising its ability to acquire new knowledge. This effective resistance to catastrophic forgetting marks a significant stride in advancing quantum continual learning.
Catastrophic forgetting describes the fact that machine learning models will likely forget the knowledge of previously learned tasks after the learning process of a new one. It is a vital problem in the continual learning scenario and recently has attracted tremendous concern across different communities. We explore the catastrophic forgetting phenomena in the context of quantum machine learning. It is found that, similar to those classical learning models based on neural networks, quantum learning systems likewise suffer from such forgetting problem in classification tasks emerging from various application scenes. We show that based on the local geometrical information in the loss function landscape of the trained model, a uniform strategy can be adapted to overcome the forgetting problem in the incremental learning setting. Our results uncover the catastrophic forgetting phenomena in quantum machine learning and offer a practical method to overcome this problem, which opens a new avenue for exploring potential quantum advantages towards continual learning.
ABSTRACT Background Quantum federated learning presents a promising paradigm for privacy‐preserving collaborative training across distributed quantum devices. However, its scalability is hindered by the significant communication overhead associated with transmitting high‐dimensional, high‐precision quantum model parameters over classical networks. Methods To address this bottleneck, this paper proposes QFI‐Opt (Quantum Fisher Information‐guided Adaptive Optimization), a quantum adaptive communication optimization framework based on Quantum Fisher Information. QFI‐Opt establishes a “sensing‐compression‐regulation” pipeline that achieves communication efficiency while preserving quantum model fidelity. The framework uses QFI as a physically interpretable metric to dynamically assess quantum state sensitivity to parameter perturbations, enabling a progressive pruning strategy that removes low‐sensitivity parameters during training while retaining critical quantum features. Additionally, a dynamic bit‐width quantization mechanism adapts precision based on parameter importance, maximizing compression without compromising numerical stability. This is further complemented by a physics‐aware aggregation method that weighs client updates based on both local data volume and quantum information quality derived from QFI scores, improving global model robustness. Results Extensive evaluation on quantum convolutional neural networks demonstrates that QFI‐Opt significantly reduces per‐round communication overhead compared to baseline methods. Conclusions Simultaneously, the proposed framework maintains competitive model accuracy and convergence performance across diverse quantum architectures.
Abstract Variational quantum algorithms are among the most widely studied approaches for applying near-term quantum devices to learning tasks. In this work we investigate a supervised quantum-autoencoder-style classifier that encodes images by amplitude encoding and predicts labels from a designated readout sub-register. Building on the quantum-autoencoder (QAE)-style classifier of Asaoka and Kudo, we make three contributions. First, we give a rigorous equivalence result showing that, for computational-basis label states, the original SWAP-test-based training objective can be estimated without ancilla and reference registers by directly measuring the readout subsystem. This simplification reduces circuit resources while preserving the same training signal. Second, we propose a lightweight hybrid extension: instead of using only the probability of the correct readout bitstring, we estimate the full readout distribution and apply a learnable temperature-and-bias post-processing followed by a softmax normalization. Across MNIST, KMNIST, and Fashion-MNIST (restricted to eight classes), this classical post-processing yields substantial and statistically significant improvements in evaluation accuracy, consistently across cross-entropy and mean-squared-error objectives. Finally, to interpret these gains we analyze Fisher-geometry summaries based on effective dimension and complementary Fisher-spectrum diagnostics for both the quantum models and the classical baselines. Empirically, the hybrid quantum model is associated with a more spectrally concentrated and anisotropic Fisher geometry, with Fisher mass concentrated in fewer dominant directions and with a substantially larger local Fisher scale after training.