Mirror subspace diagonalization: A real-time quantum Krylov algorithm with near-optimal sampling cost
summary
The gist
The mirror subspace diagonalization (MSD) algorithm is introduced as an alternative to conventional quantum Krylov algorithms, aiming to approach the theoretical lower bound of sampling cost for
In short
Mirror subspace diagonalization (MSD) is a new quantum Krylov algorithm designed to estimate ground-state energy with near-optimal sampling cost. It uses a finite-difference formula involving symmetrically shifted timesteps to efficiently estimate the Hamiltonian matrix. This method significantly reduces the required number of measurements compared to conventional methods, achieving reductions of 10 to 10,000 times for molecular simulations.
Key concepts
- Quantum Krylov Algorithms
- These are methods used on quantum computers to estimate ground-state energy by evolving a Hamiltonian operator over time. The main drawback is that they require measuring each term individually, making the sampling cost very high and slow for large problems.
- Mirror Subspace Diagonalization (MSD)
- MSD is an approach that tries to solve the high sampling cost problem. It uses a special finite-difference formula with shifted timesteps to express the Hamiltonian as a combination of time-evolution unitaries, allowing for much more efficient estimation of the relevant parts of the Hamiltonian matrix.
- Sampling Cost Lower Bound
- This is a theoretical minimum number of measurements needed for an algorithm to estimate an energy accurately. MSD aims to approach this lower bound, meaning it uses fewer measurements than standard methods while still maintaining high accuracy, especially when the Hamiltonian has specific spectral properties.
Terminology used across episodes
This episode discusses
- Mirror subspace diagonalization: A real-time quantum Krylov algorithm with near-optimal sampling cost · Paper Radio
- Quantum Filter Diagonalization: Quantum Eigendecomposition without Full Quantum Phase Estimation
- The quantum super-Krylov method
- Compilation of Trotter-Based Time Evolution for Partially Fault-Tolerant Quantum Computing Architecture
- Global Minimization of Electronic Hamiltonian 1-Norm via Linear Programming in the Block Invariant Symmetry Shift (BLISS) Method
- Krylov variational quantum algorithm for first principles materials simulations
- Quantum subspace expansion algorithm for Green's functions
- Hamiltonian simulation-based quantum-selected configuration interaction for large-scale electronic structure calculations with a quantum computer
- Quantum-selected configuration interaction with time-evolved state
- Quantum-Centric Algorithm for Sample-Based Krylov Diagonalization
The paper
Mirror subspace diagonalization: A real-time quantum Krylov algorithm with near-optimal sampling cost · Read on arXiv
Shota Kanasugi, Yuya O. Nakagawa, Norifumi Matsumoto, Yuichiro Hidaka, Kazunori Maruyama, Hirotaka Oshima
Quantum Laboratory, Fujitsu Research, Fujitsu Limited
DOI: 10.1038/s41534-026-01384-y
Transcript
Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: Today's paper: "Mirror subspace diagonalization".
Mira: The mirror subspace diagonalization (MSD) algorithm is introduced as an alternative to conventional quantum Krylov algorithms,
Kai: First, who's behind it and why it matters.
Paper summary: Kai: So, to recap, this paper introduces MSD as an alternative method that tries to get closer and closer to the theoretical minimum sampling cost for ground-state energy estimation on near-term quantum computers >
Mira: They claim this method works by leveraging a finite-difference formula to express the Hamiltonian operator in terms of time-evolution unitaries with symmetrically shifted timesteps, which lets them estimate the projected Hamiltonian matrix more efficiently >
Lev: And they do this by carefully optimizing that timestep parameter and performing a shifting of the energy spectrum, which is what they say makes it comparable to a theoretical lower bound on the sampling cost for KQD approaches >
Kai: Before we get into how they do it technically, we need to know what this actually means for us in terms of the big picture right now >
Mira: Well, the paper focuses on making sure that when you optimize that timestep and shift the spectrum, you simultaneously minimize both the sampling error and the finite-difference errors >
Lev: That's important because if you don't handle those two errors together, one might make up for itself in a way that doesn't help your overall accuracy >
Kai: And they say that by optimizing these parameters, MSD can attain that sampling cost lower bound up to a logarithmic factor >
Conclusion: Mira: So looking at the title, "Mirror subspace diagonalization: A real-time quantum Krylov algorithm with near-optimal sampling cost", it sounds like they’re really focused on making this method practical for what we can build right now >
Kai: Yeah, and the fact that they say it approaches the theoretical lower bound of sampling cost is significant because it suggests a pathway towards obtaining chemically accurate ground-state energies on near-term quantum devices >
Lev: For us building hardware, it means this algorithm could be a crucial step in reducing the total shot count you need to run these kinds of calculations >
Mira: It’s not just about running faster; it’s about finding a way to get better energy estimates without needing massive amounts of quantum resources >
Kai: So what this paper really means for the world is that we might have a concrete tool that helps us move from theoretical possibilities to actually getting accurate answers on current quantum hardware >
More episodes
- 1605.03774-Pure single photons from a trapped atom source
- 2008.03026-Heat engines with single-shot deterministic work extraction
- 2111.04378-Non-Hermitian dynamics and PT-symmetry breaking in interacting mesoscopic Rydberg platforms
- 2205.10590-Steady-state entanglement generation for non-degenerate qubits
- 2406.06530-Relativistic generalization of Feynman's path integral on the basis of extended Lagrangians
- 2410.01904-QKD as a Quantum Machine Learning task
- 2411.19601-Quantum simulation of thermal field theories
- 2506.12608-Statistical Description of Fermi System over a Surface in a Uniform External Field
- 2501.05349-Fermionic cellular automata in one dimension
- 2506.03441-Conjectured Bounds for 2-Local Hamiltonians via Token Graphs