https://github.com/cqcl/constant-scaling-quantum-gradient-estimation
https://github.com/cqcl/constant-scaling-quantum-gradient-estimation
Last synced: 8 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/cqcl/constant-scaling-quantum-gradient-estimation
- Owner: CQCL
- License: mit
- Created: 2022-11-24T17:29:28.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2024-01-19T17:27:50.000Z (over 2 years ago)
- Last Synced: 2025-02-06T21:46:22.743Z (over 1 year ago)
- Language: Jupyter Notebook
- Size: 10.3 MB
- Stars: 2
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Constant-Scaling Quantum Gradient Estimation
This repository contains the code to reproduce the results of the paper "Gradient Estimation with Constant Scaling for Hybrid Quantum Machine Learning" by Thomas Hoffmann and Douglas Brown.
# Structure
- `random_dps_test.py`: contains the experiment for random datapoints with a fully connected layer with sigmoid activation as the output layer
- `random_dps_test_wo_fc`: contains the experiment for random datapoints with the classification coming from the rescaled Z-expectation of a single qubit
- `mnist.py`: contains the experiment for the `quanvolutional` neural network based on [Henderson et al. 2019]
- `random_dps_plots`: contains the code to make Fig. 2 (results of the random datapoints experiment)
- `lr_plots.py`: contains the code to make Fig. 3 (results of investigating the effect of learning rate on the convergence of each task)
- `/data_...`: contains data for the experiment in question
- `/figs` contains figures used in the report
- `spsb_diff` and `spsb_diff_smoothing`: contains the `spsb` differentiator (with optional Jacobian smoothing) which can be imported to use in the `diff_method` argument in a PennyLane `qnode` as below:
```
import pennylane as qml
from spsb_diff import spsb_diff
dev = qml.device('default.qubit', wires=1)
@qml.qnode(dev, diff_method=spsb_diff)
def circuit(x):
qml.RX(x, wires=0)
return qml.expval(qml.PauliZ(0))
```