Reproduction Code for: On Reducing the Amount of Samples Required for Training of QNNs (doi:10.18419/darus-3445)

View:

Part 1: Document Description
Part 2: Study Description
Part 5: Other Study-Related Materials
Entire Codebook

Reproduction Code for: On Reducing the Amount of Samples Required for Training of QNNs

doi:10.18419/darus-3445

DaRUS

2023-09-27

1

Mandl, Alexander; Barzen, Johanna; Leymann, Frank; Mangold, Victoria; Riegel, Benedikt; Vietz, Daniel; Winterhalter, Felix, 2023, "Reproduction Code for: On Reducing the Amount of Samples Required for Training of QNNs", https://doi.org/10.18419/darus-3445, DaRUS, V1

Reproduction Code for: On Reducing the Amount of Samples Required for Training of QNNs

Constraints on the Linear Structure of the Training Data

doi:10.18419/darus-3445

Mandl, Alexander (University of Stuttgart, Institute of Architecture of Application Systems)

Barzen, Johanna (University of Stuttgart, Institute of Architecture of Application Systems)

Leymann, Frank (University of Stuttgart, Institute of Architecture of Application Systems)

Mangold, Victoria (University of Stuttgart, Institute of Architecture of Application Systems)

Riegel, Benedikt (University of Stuttgart, Institute of Architecture of Application Systems)

Vietz, Daniel (University of Stuttgart, Institute of Architecture of Application Systems)

Winterhalter, Felix (University of Stuttgart, Institute of Architecture of Application Systems)

01MK20005N

01MQ22007B

DaRUS

Mandl, Alexander

Mandl, Alexander

2023-05-02

https://doi.org/10.18419/darus-3445

Computer and Information Science, Quantum Neural Network, Quantum Computing, No Free Lunch Theorem, Quantum Entanglement

<p> Replication code for training Quantum Neural Networks using entangled datasets. <br />This is the version of the code that was used to generate the experiment results in the related publication. For future developments and discussion see <a href="https://github.com/UST-QuAntiL/entangled_qnn_training">the Github repository</a>. <p/> <p> <b>Experiments:</b><br /> <code>avg_rank_exp.py</code>: Experiments for training QNNs using training data of varying Schmidt rank<br /> <code>nlihx_exp.py</code>: Experiments for training QNNs using linearly dependent data<br /> <code>ortho_exp.py</code>: Experiments for training QNNs using orthogonal training data<br /> </p> <p> <b>Visualisation/Analysis of data (plots.py):</b><br /> - Generates plots for the experiments above either from the data in <code>experimental_results</code> or from the processed results (see Data).<br /> - Processes results to extract information from raw data in <code>experimental_results</code> (to change behavior see the function calls at the end of <code>plots.py</code>).<br /> </p> <p> <b>Data:</b><br /> The raw data for the experiments is available in <a href="https://doi.org/10.18419/darus-3442">the experiment dataset</a>. </p>

Github Repository: <a href="https://github.com/UST-QuAntiL/entangled_qnn_training">UST-QuAntiL/entangled_qnn_training</a>

Mandl, Alexander; Barzen, Johanna; Leymann, Frank; Mangold, Victoria; Riegel, Benedikt; Vietz, Daniel; Winterhalter, Felix, 2023, "Data Repository for: On Reducing the Amount of Samples Required for Training of QNNs", <a href="https://doi.org/10.18419/darus-3442">https://doi.org/10.18419/darus-3442</a>, DaRUS

Label:

avg_rank_exp.py

Text:

Experiments for training QNNs using training data of varying Schmidt rank.

text/x-python

Label:

classic_training.py

Text:

Cost function and training routines procedures for PyTorch QNN simulation.

text/x-python

Label:

config.py

Text:

Configuration structures for experiments.

text/x-python

Label:

cost_modifying_functions.py

Text:

Various functions to modify the loss function after evaluation.

text/x-python

Label:

data.py

Text:

Data generation routines for the experiments.

text/x-python

Label:

generate_experiments.py

Text:

Experiment entry point. Generates training data, calls simulation and training routines and saves results. Also responsible for distributing experiment workloads to multiple processors.

text/x-python

Label:

logger.py

Text:

Logging utilities

text/x-python

Label:

metrics.py

Text:

Functions for evaluating the quantum risk

text/x-python

Label:

nlihx_exp.py

Text:

Experiments for training QNNs using linearly dependent data.

text/x-python

Label:

ortho_exp.py

Text:

Experiments for training QNNs using orthogonal training data.

text/x-python

Label:

plots.py

Text:

Python script for generating the plots for the quantum risk for all experiments and for aggregating and analyzing the raw experiment results

text/x-python

Label:

README.md

Text:

Readme file with additional information.

text/markdown

Label:

requirements.txt

Text:

Python requirements for reproduction.

text/plain

Label:

utils.py

Text:

Utility functions for data generation and experiment setup.

text/x-python

Label:

vis_utils.py

Text:

Utility functions for visualisation.

text/x-python

Label:

cuda_qnn.py

Text:

QNN ansatz implementations for classical simulation using PyTorch.

text/x-python

Label:

qnn.py

Text:

General QNN description

text/x-python

Label:

quantum_gates.py

Text:

Gate implementations of common quantum gates for simulation.

text/x-python