Quantum Continual Learning Overcoming Catastrophic Forgetting

Abstract

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. In this paper, we explore the catastrophic forgetting phenomena in the context of quantum machine learning. We find 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.

References in corpus (37)

• Playing Atari with Deep Reinforcement Learning
• Quantum Machine Learning
• Experimental Observation of the Quantum Anomalous Hall Effect in a Magnetic Topological Insulator
• Hardware-efficient Variational Quantum Eigensolver for Small Molecules and Quantum Magnets
• Machine learning and the physical sciences
• Quantum machine learning in feature Hilbert spaces
• Quantum Convolutional Neural Networks
• An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networks
• Quantum Fisher information matrix and multiparameter estimation
• Quantum generative adversarial learning
• Quantum Simulators: Architectures and Opportunities
• A rigorous and robust quantum speed-up in supervised machine learning
• Quantum algorithms for supervised and unsupervised machine learning
• Development of Quantum InterConnects for Next-Generation Information Technologies
• Quantum Boltzmann Machine
• Hierarchical quantum classifiers
• The Expressive Power of Parameterized Quantum Circuits
• Information-theoretic bounds on quantum advantage in machine learning
• Quantum Computer Systems for Scientific Discovery
• Stochastic gradient descent for hybrid quantum-classical optimization
• Measuring Catastrophic Forgetting in Neural Networks
• Experimental quantum speed-up in reinforcement learning agents
• Quantum generative adversarial learning in a superconducting quantum circuit
• Quantum noise protects quantum classifiers against adversaries
• Machine learning meets quantum physics
• Introduction to quantum Fisher information
• Quantum Adversarial Machine Learning
• Statistical mechanics of the Cluster-Ising model
• Quantum Discriminant Analysis for Dimensionality Reduction and Classification
• Basic protocols in quantum reinforcement learning with superconducting circuits
• Continual Unsupervised Representation Learning
• Vulnerability of quantum classification to adversarial perturbations
• Reformulation of the No-Free-Lunch Theorem for Entangled Data Sets
• Realizing a quantum generative adversarial network using a programmable superconducting processor
• Limitations of the Empirical Fisher Approximation for Natural Gradient Descent
• Universal Adversarial Examples and Perturbations for Quantum Classifiers
• Continual Learning in Neural Networks

Cited by in corpus (4)

• Recent advances for quantum classifiers
• Quantum continual learning on a programmable superconducting processor
• Quantum continual learning of quantum data realizing knowledge backward transfer
• Universal adversarial perturbations for multiple classification tasks with quantum classifiers