Quantum Generative Training Using Rényi Divergences
arXiv:2106.09567
Abstract
Quantum neural networks (QNNs) are a framework for creating quantum algorithms that promises to combine the speedups of quantum computation with the widespread successes of machine learning. A major challenge in QNN development is a concentration of measure phenomenon known as a barren plateau that leads to exponentially small gradients for a range of QNNs models. In this work, we examine the assumptions that give rise to barren plateaus and show that an unbounded loss function can circumvent the existing no-go results. We propose a training algorithm that minimizes the maximal Rényi divergence of order two and present techniques for gradient computation. We compute the closed form of the gradients for Unitary QNNs and Quantum Boltzmann Machines and provide sufficient conditions for the absence of barren plateaus in these models. We demonstrate our approach in two use cases: thermal state learning and Hamiltonian learning. In our numerical experiments, we observed rapid convergence of our training loss function and frequently archived a average fidelity in fewer than epochs.
References in corpus (24)
- Barren plateaus in quantum neural network training landscapes
- QuTiP: An open-source Python framework for the dynamics of open quantum systems
- Rényi Divergence and Kullback-Leibler Divergence
- The Role of Relative Entropy in Quantum Information Theory
- The foundations of statistical mechanics from entanglement: Individual states vs. averages
- Circuit-centric quantum classifiers
- Noise-Induced Barren Plateaus in Variational Quantum Algorithms
- Simulating Hamiltonian dynamics with a truncated Taylor series
- Connecting ansatz expressibility to gradient magnitudes and barren plateaus
- Strong converse for the classical capacity of entanglement-breaking and Hadamard channels via a sandwiched Renyi relative entropy
- A Quantum Algorithm for Finding the Minimum
- Layerwise learning for quantum neural networks
- Robust Online Hamiltonian Learning
- Sampling from the thermal quantum Gibbs state and evaluating partition functions with a quantum computer
- Trainability of Dissipative Perceptron-Based Quantum Neural Networks
- Tomography and Generative Data Modeling via Quantum Boltzmann Training
- Large gradients via correlation in random parameterized quantum circuits
- A Quantum-Quantum Metropolis Algorithm
- Higher Order Derivatives of Quantum Neural Networks with Barren Plateaus
- A Variational Quantum Algorithm for Preparing Quantum Gibbs States
- Quantum Bootstrapping via Compressed Quantum Hamiltonian Learning
- Qubit-efficient entanglement spectroscopy using qubit resets
- Generative training of quantum Boltzmann machines with hidden units
- Smallest eigenvalue distribution of the fixed trace Laguerre beta-ensemble