Quantum Neural Networks and Computational Methods

Quantum neural networks (QNNs) represent a convergence of quantum information science and machine learning, seeking to harness quantum systems’ parallelism and entanglement to process data in ways that classical networks cannot. By encoding inputs into quantum states and employing parameterised gate sequences, QNNs aim to perform learning tasks through variational circuits that adjust quantum parameters to minimise cost functions. Computational methods span both discrete qubit models and continuous-variable frameworks, each offering distinct advantages in terms of state encoding and circuit depth. Core challenges include realising non‐linear activation functions within the constraints of unitary evolution, mitigating errors in near‐term devices and scaling architectures to practical problem sizes. Recent progress has demonstrated the potential for exponential storage capacity, enhanced pattern recognition and secure cryptographic protocols, signalling a shift towards algorithms that exploit superposition, entanglement and interference for tasks such as classification, optimisation and secure communication.

Research from Nature Portfolio

Recent studies have introduced a continuous‐variable quantum neural cryptosystem in which quantum modes encode classical and quantum data for key generation, encryption and decryption. By integrating Gaussian operations with adaptive learning rates, this framework achieves secure information transfer with demonstrated feasibility on photonic simulation platforms. Optimisation of learning parameters led to a reduction in computational time and improved decryption fidelity across a range of input data, underscoring the promise of continuous‐variable architectures for both machine learning and quantum cryptography.

Research from all publishers

In superconducting processors, repeat‐until‐success circuits have been used to implement quantum neurons that exhibit non‐linear activation by leveraging real‐time feedback. A minimal feedforward QNN assembled from these neurons successfully classified all 2‐to‐1‐bit Boolean functions, illustrating coherent deep‐learning capabilities on noisy hardware. Meanwhile, a general quantum algorithm has been developed to approximate arbitrary analytic activation functions via controlled gate sequences without intermediate measurement. This approach confers universal approximation power on feedforward QNNs, enabling gradient‐based training and extending the scope of quantum machine learning to a broader class of analytic models.

Quantum Neural Networks and Computational Methods publication trend

The graph below shows the total number of articles in quantum neural networks and computational methods across all publications each year (not limited to Nature Index journals).

Technical terms

Qubit: The fundamental two‐level quantum unit of information that can exist in a superposition of basis states.

Superposition: A property of quantum systems in which a qubit occupies multiple basis states simultaneously until measurement.

Entanglement: A correlation between quantum systems such that the state of each cannot be described independently of the state of the others.

Variational Quantum Circuit: A parameterised sequence of quantum gates optimised by a classical algorithm to minimise a cost function.

Activation Function: A non‐linear operation applied within a neural network to introduce complexity and enable learning of intricate patterns.

Continuous‐Variable Quantum System: A quantum computing model that uses bosonic modes (such as light fields) with continuous spectrum observables for information processing.

Repeat-Until-Success Circuit: A probabilistic quantum circuit that employs conditional operations and real‐time feedback to implement non‐unitary transformations.

References

1. Realization of a quantum neural network using repeat-until-success circuits in a superconducting quantum processor. npj Quantum Information (2023).
2. An Approach to Cryptography Based on Continuous-Variable Quantum Neural Network. Scientific Reports (2020).
3. Quantum activation functions for quantum neural networks. Quantum Information Processing (2022).
4. Storage capacity and learning capability of quantum neural networks. Quantum Science and Technology (2021).