Trainability of Quantum Neural Networks: Barren Plateaus and Scalability

Marco Cerezo de la Roca, Akira Sone, Kunal Sharma +3 more

Han Qi, Lei Wang, Hongsheng Zhu +2 more

Ruipeng Xing, Yanan Li, Zhen Shang +3 more

Barren plateaus -- the exponential vanishing of gradients -- are a fundamental obstacle to training scalable quantum neural networks. Whether they arise in quantum recurrent neural networks (QRNNs), a natural architecture for sequential data, remains a pressing question. Here we show that the decisive ingredient for trainability in QRNNs is not the recurrent circuit topology per se, but enforcing time-translation symmetry through parameter sharing across time steps. We prove that, without parameter sharing, QRNNs suffer from barren plateaus, with gradient variance decaying exponentially with sequence length. Imposing parameter sharing across time steps fundamentally alters this scaling, transforming it into a polynomial dependence and thereby suppressing the barren plateau. Numerical simulations corroborate these analytical predictions. By rigorously showing how time-translation symmetry suppresses barren plateaus and enhances learning capability in QRNNs, our work establishes task-aligned symmetry as a constructive resolution to the expressivity-trainability tension in quantum neural networks.

Jacob L. Cybulski, Thanh Thi Nguyen

Kunal Sharma, Marco Cerezo, Łukasz Cincio +1 more

Several architectures have been proposed for quantum neural networks (QNNs), with the goal of efficiently performing machine learning tasks on quantum data. Rigorous scaling results are urgently needed for specific QNN constructions to understand which, if any, will be trainable at a large scale. Here, we analyze the gradient scaling (and hence the trainability) for a recently proposed architecture that we call dissipative QNNs (DQNNs), where the input qubits of each layer are discarded at the layer's output. We find that DQNNs can exhibit barren plateaus, i.e., gradients that vanish exponentially in the number of qubits. Moreover, we provide quantitative bounds on the scaling of the gradient for DQNNs under different conditions, such as different cost functions and circuit depths, and show that trainability is not always guaranteed. Our work represents the first rigorous analysis of the scalability of a perceptron-based QNN.

Arthur Chalom Pesah, Marco Cerezo, Samson Wang +3 more

Quantum neural networks (QNNs) have generated excitement around the possibility of efficiently analyzing quantum data.But this excitement has been tempered by the existence of exponentially vanishing gradients, known as barren plateau landscapes, for many QNN architectures.Recently, quantum convolutional neural networks (QCNNs) have been proposed, involving a sequence of convolutional and pooling layers that reduce the number of qubits while preserving information about relevant data features.In this work, we rigorously analyze the gradient scaling for the parameters in the QCNN architecture.We find that the variance of the gradient vanishes no faster than polynomially, implying that QCNNs do not exhibit barren plateaus.This result provides an analytical guarantee for the trainability of randomly initialized QCNNs, which highlights QCNNs as being trainable under random initialization unlike many other QNN architectures.To derive our results, we introduce a novel graph-based method to analyze expectation values over Haar-distributed unitaries, which will likely be useful in other contexts.Finally, we perform numerical simulations to verify our analytical results.