Analysis notebooks for report

CQCL / qbm_benchmark_dataset

A simple python package to benchmark Quantum Boltzmann Machine models using Stochastic Gradient Descent. Based on the quantum many-body physics package `quimb`

Apache License 2.0

1 stars 0 forks source link

Analysis notebooks for report #27

Closed erinaldiq closed 6 months ago

erinaldiq commented 6 months ago

The newly added notebooks are used to investigate the results and create plots used for analysis.

panasonic-sakata-ryuji commented 6 months ago

I would like to confirm a few points.

Regarding 1d_heisenberg, what is the reason for using results with a learning_rate of 0.025 in the analysis notebook?
Would it be possible to add a discussion on the effects of noise and pre-training?

yutakikuchiQ commented 6 months ago

Hi @panasonic-sakata-ryuji san,

We initially tested the Heisenberg model with the learning_rate=0.25, but we numerically found the training failing. Therefore, we decreased the learning rate to 0.025 to successfully train the model.
You can find an extensive discussion on the effect of noise and pre-training in the final report. However, it is indeed a good idea to have the corresponding discussion in the notebooks and you are welcome to edit them.