"From zero to hero" Qiskit Machine Learning tutorial on a real dataset
Opened this issue · 6 comments
Description
Quantum machine learning is gaining momentum and becoming more and more popular across machine learning researchers. This project aims facilitating smooth transition from classical to quantum machine learning algorithms for those who are already proficient in classical ML. The goal of this project is to pick up a small classical machine learning dataset and an ML problem, e.g. classification, regression, then build up a quantum model that will produce similar results. The dataset should be small enough to be processed by a quantum computer or a simulator, but in the same time it should be widely known by ML practitioners. A classical iris dataset may be a good candidate. The tutorial itself should be as close as possible to a real life Qiskit Machine Learning use case. This tutorial should be a good starting point for everybody.
Deliverables
A tutorial, jupyter notebook in Qiskit Machine Learning repository, explaining how to move from a classical algorithm an analogous quantum one.
Mentors details
Anton Dekusar, @adekusar-drl
Research Software Engineer / Qiskit Machine Learning contributor
Number of mentees
1
Type of mentees
What is required:
- Excellent writing skills and proficiency in jupyter notebook
- Proficiency in classic machine learning algorithms
- Basic knowledge of quantum machine learning
I al very interested to join the team !!!
I hereby do an upload of my intermediate reporting on 7 April 2022.
QAMP Spring 2022 - #5 From zero to hero - Eric Michiels - April 2022.pptx
Checkpoint on 5-May-2022
My project consists of 5 phases. I have been mentored very professionally during all of them.
In the first phase I analyzed two existing Quantum Machine Learning (QML) tutorials that are available in Qiskit. The drive to understand each piece of code to the ultimate detail allowed me to learn a lot. I also updated my Qiskit QML version up to the most recent level.
In the second phase, my mentor introduced me to a popular dataset that is used often in Classical Machine Learning (CML) tutorials and examples, the Iris Dataset. I applied several analysis techniques to understand its structure and content. While the Iris Dataset is relatively simple, being a set of 150 rows where each row maps 4 features of an Iris flower to a specific Iris species, the principles can be applied and reused for other datasets. By the way, there are only 3 species of Iris flowers in the dataset represented, with 50 entries for each species.
In phase 3, I used CML models to predict the species of Iris flowers based on 4 input features. Training and Test Datasets where used to create a Model and this was applied on new features. The features are sepal length, sepal width, petal length and petal width.
And now phase 4 needs to be started. In this phase I will apply Quantum Neural Network techniques on the Iris Dataset and compare them with the results of phase 3.
In phase 5 I will bring all work together in a final tutorial, in which I also will refer to alternative Datasets for the Iris Dataset, consisting of more features and data.
The attached visualization contains the flowers my QML Model must be able to classify.
Delivery Presentation and Jupyter Notebook for 9-Jun-2022
#5 From zero to hero - QML tutorial on real dataset - Eric Michiels - June 2022.pdf
@HuangJunye Please find my deliverables in this issue :-)
Eric Michiels.