Critical Points in Quantum Generative Models

Authors: Eric Ricardo Anschuetz

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We now test our analytic predictions using numerical simulations.
Researcher Affiliation Academia Eric R. Anschuetz MIT Center for Theoretical Physics Cambridge, MA 02139, USA eans@mit.edu
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide a specific link or explicit statement about the availability of its source code.
Open Datasets No The paper uses a Hamiltonian and generates problem instances for simulation, rather than using a publicly available or open dataset for which access information is provided. "HT,U is the 1D n site spinless Fermi Hubbard Hamiltonian (Negele & Orland, 1998) at half filling. Here, we take units such that the mean eigenvalue of the considered Hamiltonian (minus its smallest eigenvalue) is E = 1."
Dataset Splits No The paper does not explicitly provide training, validation, or test dataset splits. The numerical experiments describe generating instances for simulation: "To estimate the empirical distribution of local minima for the studied instances of the varitional quantum eigensolver (VQE) (Peruzzo et al., 2014), we repeated this procedure 52 times, using a new ansatz and uniformly random starting point for each training instance."
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used to run its experiments.
Software Dependencies No The paper mentions software used, 'Qiskit (Abraham et al., 2019)', but does not provide specific version numbers for it or any other ancillary software dependencies.
Experiment Setup Yes Our implementation of gradient descent used a learning rate of 0.05 and a momentum of 0.9, and halted when either the function value improved by no more than 10 5 or after 106 iterations, whichever came first. We initialized each instance at a uniformly random point in parameter space, with each parameter initialized within [ 2π, 2π].