Python package for visualizing the loss landscape of parameterized quantum algorithms.

Last update: Dec 30, 2022

Related tags

Overview

orqviz

A Python package for easily visualizing the loss landscape of Variational Quantum Algorithms by Zapata Computing Inc.

orqviz provides a collection of tools which quantum researchers and enthusiasts alike can use for their simulations. It works with any framework for running quantum circuits, for example qiskit, cirq, pennylane, and Orquestra. The package contains functions to generate data, as well as a range of flexible plotting and helper functions. orqviz is light-weight and has very few dependencies.

Getting started

In doc/examples/ we provide a range of Jupyter notebook examples for orqviz. We have four Jupyter notebooks with tutorials for how to get started with any quantum circuit simulation framework you might use. You will find examples with qiskit, cirq, pennylane and Zapata's Orquestra library. The tutorials are not exhaustive, but they do provide a full story that you can follow along.

In this notebook we have the Sombrero example that we showcase in our paper. We also have an advanced example notebook which provides a thorough demonstration of the flexibility of the orqviz package.

We recently published a paper on arXiv where we review the tools available with orqviz:
ORQVIZ: Visualizing High-Dimensional Landscapes in Variational Quantum Algorithms

Installation

You can install our package using the following command:

pip install orqviz

Alternatively you can build the package from source. This is especially helpful if you would like to contribute to orqviz

git clone https://github.com/zapatacomputing/orqviz.git
cd orqviz
pip install -e ./

Examples

import orqviz
import numpy as np

np.random.seed(42)

def loss_function(pars):
    return np.sum(np.cos(pars))**2 + np.sum(np.sin(30*pars))**2

n_params = 42
params = np.random.uniform(-np.pi, np.pi, size=n_params)
dir1 = orqviz.geometric.get_random_normal_vector(n_params)
dir2 = orqviz.geometric.get_random_orthonormal_vector(dir1)

scan2D_result = orqviz.scans.perform_2D_scan(params, loss_function,
                                direction_x=dir1, direction_y=dir2,
                                n_steps_x=100)
orqviz.scans.plot_2D_scan_result(scan2D_result)

This code results in the following plot:

Authors

The leading developer of this package is Manuel Rudolph at Zapata Computing.
For questions related to the visualization techniques, contact Manuel via [email protected] .

The leading software developer of this package is Michał Stęchły at Zapata Computing.
For questions related to technicalities of the package, contact Michał via [email protected] .

Thank you to Sukin Sim and Luis Serrano from Zapata Computing for their contributions to the tutorials.

You can also contact us or ask general questions using GitHub Discussions.

For more specific code issues, bug fixes, etc. please open a GitHub issue in the orqviz repository.

If you are doing research using orqviz, please cite our paper:

ORQVIZ: Visualizing High-Dimensional Landscapes in Variational Quantum Algorithms

How to contribute

Please see our Contribution Guidelines.

Comments

Use transpile to build circuit only once

Despite being wrapped up in a lambda function, the get_circuit function is actually still called for every function evaluation during plot generation or optimization, and hence the circuit is rebuilt each time. This rather defeats the concept of late binding of the parameter values. The PR uses a slightly different approach using the transpile function. The code is arguably more transparent than using the lambda function wrapper. Evaluation is faster now, but for this simple case rarely more than 10%. One downside, the circuit cannot be plotted anymore in a simple way.

opened by RonMeiburg 11
ci: add step with Pythonic cruft cleanup

Apparently, issues that we had with mypy stem from Github Actions caching some (?) directories (thanks @alexjuda for pointing this out!). This PR adds a cleaning step (taken from z-quantum-actions) that deletes potentially conflicting directories.

opened by dexter2206 1
Clean notebooks

These are the once-run versions of the cirq and qiskit notebooks derived from the 'simplified qiskit get_circuit() return' commit from the main branch. I hope this works for you. If not, then I apologize, When it comes to git I still suffer from sas every now and then.

opened by RonMeiburg 1
Loss function clarity
Goals of this draft PR:

Allow parameters to be any np.ndarray rather than strictly a 1D np.ndarray

Improve docstrings for what we define as a loss function

Improve README to specify what we define as a loss function, and how to wrap their loss function with functools.partial

Alternatively, allow loss_function_kwargs in the scanning functions that we pass to the loss_function with more than one argument.
opened by MSRudolph 1
Utilize tqdm progress bar when verbose=True during scans.
Is your feature request related to a problem? Please describe. We should replace the print calls when verbose=True in scans with tqdm from the tqdm library. Alternatively, we make it the default and find a way to mute the library's prints.

Describe the solution you'd like

verbose = True # or False for it in tqdm(range(n_iters), disable = not verbose): ... # run scans

Additional context Our verbosity options are currently very rudimentary and tqdm is one of the most used Python libraries.
opened by MSRudolph 2
This repo should contain Issues for how people can contribute

Is your feature request related to a problem? Please describe. Currently, when people enter the orqviz GitHub repository with the intent to contribute, there are no open Issues and not many PRs. They will not know what might be low-hanging fruit to contribute.

Describe the solution you'd like We (orqviz developers) should open Issues which are connected to how people can concretely contribute. For example, we could provide links to existing tutorials which we believe can be readily enhanced with our visualization techniques. In such cases, potential contributors could work on such enhancements and reach out to the authors of the original tutorials. Similarly, we can elaborate on future visualization techniques which we could experiment with. This may be done by external contributors.

opened by MSRudolph 0

Releases(v0.3.0)

v0.3.0(Aug 19, 2022)
What's Changed

ci: add step with Pythonic cruft cleanup by @dexter2206 in https://github.com/zapatacomputing/orqviz/pull/43

Update main by @mstechly in https://github.com/zapatacomputing/orqviz/pull/44

Fourier transform by @laurgao in https://github.com/zapatacomputing/orqviz/pull/45

Full Changelog: https://github.com/zapatacomputing/orqviz/compare/v0.2.0...v0.3.0
Source code(tar.gz)
Source code(zip)
orqviz-0.3.0-py3-none-any.whl(38.09 KB)
v0.2.0(Feb 8, 2022)
New features:

orqviz now doesn't require parameters to be 1D vectors, they can be ND arrays instead

We introduced LossFunctionWrapper as a utility tool that helps with changing arbitrary python functions into orqviz-compatible loss functions.

Minor changes:

Improvements in notebook examples

Improvements in readme and contribution guidelines

Source code(tar.gz)
Source code(zip)
v0.1.1(Nov 9, 2021)
What's Changed

Fixed classifiers in the setup.cfg

Minor fixes in in README

Relax dependency versions

Full Changelog: https://github.com/zapatacomputing/orqviz/compare/v0.1.0...v0.1.1
Source code(tar.gz)
Source code(zip)
v0.1.0(Nov 9, 2021)

Source code(tar.gz)
Source code(zip)

Owner

Zapata Computing, Inc.

GitHub Repository

Predicting path with preference based on user demonstration using Maximum Entropy Deep Inverse Reinforcement Learning in a continuous environment

Preference-Planning-Deep-IRL Introduction Check my portfolio post Dependencies Gym stable-baselines3 PyTorch Usage Take Demonstration python3 record.

9 Oct 26, 2022

FCOS: Fully Convolutional One-Stage Object Detection (ICCV'19)

FCOS: Fully Convolutional One-Stage Object Detection This project hosts the code for implementing the FCOS algorithm for object detection, as presente

3.1k Jan 05, 2023

A toolkit for document-level event extraction, containing some SOTA model implementations

❤️ A Toolkit for Document-level Event Extraction with & without Triggers Hi, there 👋 . Thanks for your stay in this repo. This project aims at buildi

159 Dec 22, 2022

Technical Indicators implemented in Python only using Numpy-Pandas as Magic - Very Very Fast! Very tiny! Stock Market Financial Technical Analysis Python library . Quant Trading automation or cryptocoin exchange

MyTT Technical Indicators implemented in Python only using Numpy-Pandas as Magic - Very Very Fast! to Stock Market Financial Technical Analysis Python

34 Dec 27, 2022

Python package for visualizing the loss landscape of parameterized quantum algorithms.

Related tags

Overview

orqviz

Getting started

Installation

Examples

Authors

How to contribute

Comments

Use transpile to build circuit only once

ci: add step with Pythonic cruft cleanup

Clean notebooks

Loss function clarity

Utilize tqdm progress bar when verbose=True during scans.

This repo should contain Issues for how people can contribute

Releases(v0.3.0)

v0.3.0(Aug 19, 2022)

What's Changed

v0.2.0(Feb 8, 2022)

v0.1.1(Nov 9, 2021)

What's Changed

v0.1.0(Nov 9, 2021)

Owner

Zapata Computing, Inc.

Predicting path with preference based on user demonstration using Maximum Entropy Deep Inverse Reinforcement Learning in a continuous environment

FCOS: Fully Convolutional One-Stage Object Detection (ICCV'19)

A toolkit for document-level event extraction, containing some SOTA model implementations

Technical Indicators implemented in Python only using Numpy-Pandas as Magic - Very Very Fast! Very tiny! Stock Market Financial Technical Analysis Python library . Quant Trading automation or cryptocoin exchange

Code for our ICASSP 2021 paper: SA-Net: Shuffle Attention for Deep Convolutional Neural Networks

Implementation of Segnet, FCN, UNet , PSPNet and other models in Keras.

Lexical Substitution Framework

Code for "Learning Skeletal Graph Neural Networks for Hard 3D Pose Estimation" ICCV'21

Self-supervised Point Cloud Prediction Using 3D Spatio-temporal Convolutional Networks

这是一个利用facenet和retinaface实现人脸识别的库，可以进行在线的人脸识别。

Lua-parser-lark - An out-of-box Lua parser written in Lark

Repository of 3D Object Detection with Pointformer (CVPR2021)

[NeurIPS 2021] Introspective Distillation for Robust Question Answering

Weakly supervised medical named entity classification

Official MegEngine implementation of CREStereo(CVPR 2022 Oral).

The mini-MusicNet dataset

FIGARO: Generating Symbolic Music with Fine-Grained Artistic Control

NaijaSenti is an open-source sentiment and emotion corpora for four major Nigerian languages

Hierarchical Memory Matching Network for Video Object Segmentation (ICCV 2021)

A smaller subset of 10 easily classified classes from Imagenet, and a little more French