Estimating the ground-state energy of complex systems is a fundamental task in quantum physics, with wide-ranging applications in materials science, condensed-matter physics, and quantum chemistry. Despite its importance, this problem is computationally difficult, even for quantum computers. Researchers have proposed various approaches to tackle this challenge, including algorithms based on quantum phase estimation and hybrid quantum-classical optimizers involving parameterized quantum circuits. However, these methods have limitations, such as requiring large-depth quantum circuits or suffering from the “barren-plateau problem.”
Recently, a team of researchers led by Dhrumil Patel, Daniel Koch, Saahil Patel, and Mark M. Wilde has explored an alternative approach based on quantum Boltzmann machines (QBMs). QBMs use parameterized thermal states to estimate ground-state energies, and they appear to be viable due to recent progress in preparing thermal states on quantum computers. The team’s work provides a rigorous mathematical proof that the QBM learning approach is sample efficient, overcoming a key obstacle to efficient training of QBMs. This breakthrough has significant implications for the development of practical quantum algorithms for estimating ground-state energies.
Quantum Boltzmann Machines for Ground-State Energy Estimation
Estimating the ground-state energy of Hamiltonians is a fundamental task in quantum physics, with wide-ranging applications in materials science, condensed-matter physics, and quantum chemistry. Recently, researchers have proposed various approaches to tackle this problem using classical computers and quantum computers. One such approach involves employing Quantum Boltzmann Machines (QBMs), which substitute parameterized quantum circuits with parameterized thermal states of a given Hamiltonian.
Background: Variational Principle and Quantum Phase Estimation
The variational principle is one of the oldest and most widely used approaches for calculating ground-state energies on classical computers. This method reduces the search space by parameterizing a family of trial ground states and then searches over this reduced space using gradient-descent-like algorithms. Matrix product states are a powerful example of such methods, which perform well in practice.
In another direction, researchers have argued that quantum computers could be effective at calculating ground-state energies due to their ability to simulate quantum mechanical processes faithfully and with reduced overhead compared to classical algorithms. One of the first approaches proposed for doing so involves employing the quantum phase estimation algorithm for small molecules. More recently, other phase-estimation-based algorithms for ground-state energy estimation have been proposed and analyzed.
Variational Quantum Eigensolver (VQE) Approach
The VQE approach employs parameterized quantum circuits (PQCs) of shorter depth and involves a hybrid interaction between such shorter-depth quantum circuits and a classical optimizer. This approach provides a quantum computational implementation of the variational method. However, later research indicated several bottlenecks associated with VQE, including the barren-plateau problem, which will likely preclude VQE from achieving practical quantum advantage in the near term.
Quantum Boltzmann Machines (QBMs) Approach
In contrast, QBMs involve using parameterized thermal states of a given Hamiltonian and performing the search over parameterized thermal states. This approach appears to be viable due to significant recent progress on preparing thermal states on quantum computers, despite known worst-case complexity-theoretic barriers. QBMs have been analyzed in the context of Hamiltonian learning and generative modeling, but not yet for ground-state energy estimation.
Main Results: Sample Efficiency and Overcoming Barren Plateau Problem
The main finding of this paper is a rigorous mathematical proof that the QBM learning approach to approximating ground-state energies is sample efficient. The number of samples of parameterized thermal states used by our algorithm is polynomial in several quantities of interest, including the dimensionality of the Hilbert space, the number of parameters, and the desired accuracy.
In doing so, we also overcome a key obstacle to efficient training of QBMs, which was previously discussed. Our results provide analytical evidence that visible QBMs do not suffer from the barren-plateau problem, making them a promising approach for ground-state energy estimation.
Implications and Future Directions
The QBM approach offers a viable alternative to VQE for ground-state energy estimation, with potential advantages in terms of sample efficiency and robustness against barren plateaus. Further research is needed to explore the applicability of QBMs to larger systems and more complex Hamiltonians, as well as to develop practical implementations on near-term quantum devices.
External Link: Click Here For More
