Quantum Boltzmann Machine Training via Ancilla Thermometer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Quantum Boltzmann machines face challenges in training due to the NP-hard problem of sampling from quantum thermal distributions, which limits their effectiveness on near-term quantum computing devices.
Innovation Solution
The approach involves preparing a quantum Boltzmann machine in a pure state and evolving it with a chaotic, tunable quantum Hamiltonian to locally approximate a quantum thermal state, allowing for efficient sampling of observables and estimation of the inverse temperature for gradient-based training methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If gradient-based training methods are used for quantum Boltzmann machines, then training capability is improved, but sampling from quantum thermal distributions becomes NP-hard
Solution Approach 1:
The patent introduces an ancilla system (thermometer) as an intermediary to measure the temperature of the quantum Boltzmann machine. The ancilla system couples to the QBM and allows temperature estimation through local measurements, avoiding the need for direct sampling from the full quantum thermal distribution. This mediator enables gradient-based training while circumventing the NP-hard sampling problem.
Solution Approach 2:
The patent divides the quantum system into two parts: the quantum Boltzmann machine system and an ancilla thermometer system. By segmenting the measurement task, the patent performs local measurements on the ancilla system rather than requiring global sampling from the entire QBM thermal state. This segmentation reduces the computational complexity from NP-hard to a tractable problem.
2Reliability
If quantum annealing devices are used to prepare thermal states, then thermal state preparation is improved, but utility is limited by noise, connectivity, and coupling form
Solution Approach 1:
The patent uses an ancilla thermometer system as a mediator that can be coupled to various quantum systems regardless of their specific noise characteristics or connectivity constraints. The thermometer universally measures temperature by coupling to the system of interest, making the approach adaptable to different quantum hardware platforms including gate-model quantum computers, quantum annealers, and other near-term devices.
Solution Approach 2:
The patent changes the measurement parameter from requiring full thermal state sampling to measuring local expectation values of the ancilla system. By changing what is being measured (from global thermal properties to local ancilla observables), the method becomes applicable to noisy intermediate-scale quantum devices that cannot perform full thermal sampling but can execute local measurements.
3Measurement precision
If full quantum thermal state sampling is performed, then accurate temperature measurement is improved, but computational cost becomes prohibitive
Solution Approach 1:
The patent extracts the temperature measurement function from the complex quantum thermal state and transfers it to a simpler ancilla system. Instead of measuring temperature directly from the full QBM thermal distribution (which requires NP-hard sampling), the temperature information is extracted through the ancilla's local observables, which can be measured efficiently.
Solution Approach 2:
The patent creates a copy of the temperature measurement function in the ancilla system. The ancilla is prepared in a state that encodes temperature information through its coupling to the QBM, and this copied temperature information can be read out through simple local measurements rather than requiring access to the full thermal state.
Data Source
AI summary
A hybrid quantum-classical (HQC) computer prepares a quantum Boltzmann machine (QBM) in a pure state. The state is evolved in time according to a chaotic, tunable quantum Hamiltonian. The pure state locally approximates a (potentially highly correlated) quantum thermal state at a known temperature. With the chaotic quantum Hamiltonian, a quantum quench can be performed to locally sample observables in quantum thermal states. With the samples, an inverse temperature of the QBM can be approximated, as needed for determining the correct sign and magnitude of the gradient of a loss function of the QBM.