Submission Number: 100
Submission ID: 146
Submission UUID: 296d87d2-acac-43b6-ac9a-ef0b468251d9
Submission URI: /form/project

Created: Sat, 05/08/2021 - 11:05
Completed: Sat, 05/08/2021 - 12:10
Changed: Tue, 08/02/2022 - 15:00

Remote IP address: 148.74.90.25
Submitted by: Avery Leider
Language: English

Is draft: No
Webform: Project
High Performance Computing vs Quantum Computing for Neural Networks supporting Artificial Intelligence
CAREERS
Careers-image.jpg
ai (271), big-data (4), deep-learning (303), github (490), hpc-cluster-architecture (391), hpc-operations (43), jupyterhub (214), machine-learning (272), natural-language-processing (274), proposal-development (341), python (69), quantum-mechanics (281), research-facilitation (442), research-grants (447), resources (16), scikit-learn (273), singularity (23), vectorization (224)
Complete

Project Leader

Jonathan Hill
732-619-3077
212-346-1864

Project Personnel

Avery Leider, Neil McGlohon
Gio Abou Jaoude
{Empty}

Project Information

A personalized learning system that adapts to learners' interests, needs, prior knowledge, and available resources is possible with artificial intelligence (AI) that utilizes natural language processing in neural networks. These deep learning neural networks can run on high performance computers (HPC) or on quantum computers (QC). Both HPC and QC are emergent technologies. Understanding both systems well enough to select which is more effective for a deep learning AI program, and show that understanding through example, is the ultimate goal of this project. The entry to learning technologies such as HPC and QC is narrow at present because it relies on classical education methods and mentoring. The gap between the knowledge workers needed, which is in high demand, and those with the expertise to teach, which is being achieved at a much slower rate, is widening. Here, an AI cognitive agent, trained via deep learning neural networks, can help in emergent technology subjects by assisting the instructor-learner pair with adaptive wisdom. We are building the foundations for this AI cognitive agent in this project.

The role of the student facilitator will involve optimizing a deep learning neural network, comparing and contrasting with the newest technologies, such as a quantum computer (and/or a quantum computer simulator) and a high performance computer and showing the efficiency of the different computing approaches. The student facilitator will perform these tasks at the rate described in the proposal. Milestone work will be displayed and shared publicly via posting to the Jupyter Notebooks on Google Colab and linked to regular Github uploads.

Project Information Subsection

The initial Deliverable will be two Jupyter Notebooks: the first one as our validation test that we use to test our subsequent efforts, as a classical deep learning neural network posted on Google Colab and shared with a public link for viewing. The second one will be running a Deep Learning Neural Network on a real (not simulated) Quantum Computer. The code will be downloaded bare-bones Python and runnable on the High Performance Computer. Both programs will use the same Iris dataset from the scikit-learn community of machine learning researchers using Python.

The notebooks will be displayed on Google Colaboratory at: https://colab.research.google.com

This preexisting dataset is located at: https://scikit-learn.org/stable/auto_examples/datasets/plot_iris_dataset.html#sphx-glr-auto-examples-datasets-plot-iris-dataset-py

The goal is to build a quantum machine learning classifier that can be used as a foundation to a cognitive individualized tutor for general artificial intelligence. After proving the classifier works with visualizations on the Iris dataset, two other datasets will be explored also from SciKit Learn: the Wine dataset and the Breast Cancer images dataset.

The Third Deliverable will be to run the classifier on the AiMOS high performance computer and read the output results with a data plot visualization program, and that data plot visualization program also be posted on Google Colab in view mode for public consumption.

The fourth Deliverable will be to co-author with mentor a technical paper to be submitted to an outside conference to document our research.











Results will be shared with the worldwide community of researchers using Jupyter Notebooks on Google Colab and in submitting technical papers for publication in a conference.

{Empty}
A candidate I am interested in is: Gio Giorgio Abou Jaoude. His keen interest is an asset.
{Empty}
Some hands-on experience
{Empty}
Pace University
861 Bedford Rd
Pleasantville, New York. 10570
CR-Rensselaer Polytechnic Institute
06/01/2021
No
Already behind3Start date is flexible
7 months
{Empty}
06/18/2021
12/17/2021
  • Milestone Title: Display neural networks on HPC and QC
    Milestone Description: The initial Deliverable will be two Jupyter Notebooks: the first one as our validation test that we use to test our subsequent efforts, as a classical deep learning neural network posted on Google Colab and shared with a public link for viewing. The second one will be running a Deep Learning Neural Network on a real (not simulated) Quantum Computer. The code will be downloaded bare-bones Python and runnable on the High Performance Computer. Both programs will use the same Iris dataset from the scikit-learn community of machine learning researchers using Python.

    The notebooks will be displayed on Google Colaboratory at: https://colab.research.google.com

    This preexisting dataset is located at: https://scikit-learn.org/stable/auto_examples/datasets/plot_iris_dataset.html#sphx-glr-auto-examples-datasets-plot-iris-dataset-py
    Completion Date Goal: 2021-07-16
    Actual Completion Date: 2021-07-16
  • Milestone Title: Develop Quantum ML Classifier algorithm
    Milestone Description: Building a dataset is unneeded; datasets available in SciKit Learn are sufficient. Iris dataset was used for constructing the algorithm for the Quantum ML classifier. the code works for the quantum computing machine learning classifier.

    Two other datasets to test the use of the classifier are suitable more for the HPC than Google Colab:
    Wine (13 characteristic features, 3 classes)
    https://scikit-learn.org/stable/modules/generated/sklearn.datasets.load_wine.html?highlight=wine
    Breast Cancer Images (30 characteristic features, 2 classes)
    https://scikit-learn.org/stable/modules/generated/sklearn.datasets.load_breast_cancer.html?highlight=breast%20cancer

    Completion Date Goal: 2021-08-20
    Actual Completion Date: 2021-08-12
  • Milestone Title: Explain the Quantum ML Classifier
    Milestone Description: Explain in two Google Colab Jupyter Notebooks how the Quantum ML Classifier works, open-box, and also how the test validation benchmark works of the Classical deep learning Neural Network its results were compared against, also in open-box
    Completion Date Goal: 2021-09-10
    Actual Completion Date: 2021-09-01
  • Milestone Title: Write a technical paper and submit to outside conference
    Milestone Description: Final milestone will be to assist in co-authoring technical paper(s) to submit for publication in an outside conference(s) to share with the worldwide community of researchers our findings, including posting in our References the Jupyter Notebooks on Google Colab.

    Completion Date Goal: 2021-12-17
    Actual Completion Date: 2021-10-01
https://github.com/AveryLeider/QNN
Displays on Google Colab of each step accomplished on our journey to a cognitive agent using Natural Language Processing, Deep Learning Neural Networks, Quantum Computer and High Performance Computer.
(technical paper and poster) https://qce.quantum.ieee.org/authors/submission-deadlines
(technical paper and poster) https://q-se.github.io/qse2021/
(technical paper and poster) https://saiconference.com/Computing2021/CallforPapers
(ACM Conference with technical paper for Quantum Software Development) https://pldi21.sigplan.org/
(technical paper and poster) https://www.mcqst.de/news-and-events/eacn-2021/
(attendance planning: only invited papers are presented) https://qipconference.org/
They will learn how to write a research paper in IEEE format ready to submit for publication in an outside conference using LaTeX markup language favored by publishers. They will learn how to display and share their work in Google Colab and using Jupyter Notebooks. They will learn the basics of quantum computing and high performance computers that form the corpus to the natural language processing used in this effort.
{Empty}
How quantum computers fit into our future. This emergent technology is a few years away from being useful, however, the development is accelerating as heavy investment in quantum computing worldwide seeks to exploit its advantages for faster pharmaceuticals, cryptography, and military uses.
Yes, but we currently need help learning how to use HPC. We hope by running our deep learning neural network and natural language processing code on both HPC and QC will help us understand how one makes the decision that a HPC is more efficient than a QC to run a specific program.
{Empty}

Final Report

The principal disciplines of the project were (1) quantum computing and (2) machine learning, a subset of Artificial Intelligence. This project research led us into the furthest out on the edge of quantum computing, which is using quantum computation for machine learning and the latest in research on the parameter shift rule necessary for tuning the weights in back propagation for deep learning. This turned out to be possible because of the capabilities of the AiMOS supercomputer to effectively simulate quantum computing by doing the mathematics with ease.
This mostly impacts the discipline of mathematics and computer science, with some impact on quantum physics, which is the "hardware" of quantum computing - quantum bits, known as "qubits" are subatomic particles of pure energy, or sometimes very small atoms, used for the engineering of the calculations. The mathematics are all dealing with vector spaces, using linear algebra and especially matrices. The computer science subtopic is artificial intelligence, specifically machine learning, and more specifically, deep learning, as the neural networks did both forward and backward propagation, with learning algorithms on unlabeled data.
Yes, Pace University does not have a working supercomputer, so to be able to utilize the RPI's high performance computer, the AiMOS, through this program was an opportunity that would not have existed without the CAREERS program making the introductions possible for us and the availability of computer time on the AiMOS, which is the largest supercomputer in New York State.
Both the mentor, myself, and the student facilitator, Gio, learned a great deal about the steps necessary to present a program to a supercomputer for queuing in the batch, preparing it for running, running it after staging it in the right place, and then getting the output of results and concurrent output of any errors. We had to learn about placing the appropriate supporting software packages on the super computer from approved white list sources and then accessing them with the right security permissions on the work folders. We learned a great deal about supercomputing.
Pace university faculty are following our progress with great interest and asking how to also present topics to the steering committee at CAREERS for potential inclusion. The Career Services office for the Seidenberg School of Computer Science and Information Systems at Pace University has a Handshake software program and they will be posting there in their Internships section the job postings for student facilitators for projects that are posted as Recruiting on the CAREERS portal for other Pace University students to learn to about the opportunity to become student facilitators. So a relationship has begun from this one project, that hopefully will lead to more projects proposed and more skills building.
Pace university did purchase a supercomputer but did not get it running, some years ago, before the pandemic. We lost the faculty who were its advocates both at our computer science school and our business school - it was a partnership purchase between both schools. There is now great new interest in getting it running effectively. This is not easy and requires buy in from the younger faculty, but the Dean is very pleased at the new interest.
The software and the steps necessary to run our programs on the AiMOS are universal and used on high performance computers worldwide. This means that the skills learned during this experience are totally transferable to new situations. The student facilitator shared his AiMOS experience during his job interviews, and successfully got a job offer, to start at his graduation in May 2022, at TD Bank in the FinTech industry because he has had this introductory experience with a HPC from the CAREERS project. So the student facilitator is ecstatic, as even though our project had nothing to do with FinTech, it did use a HPC, which is a technology transfer of the best kind, supporting capital investments.
Yes, we have a isolated tower situation between schools. One school does not know what the other is doing with much awareness, and there is waste in duplicate areas that could be overcome by better sharing of information. However, we are in a competitive environment, so there is reticence in that. The CAREERS program causes collaboration and sharing by the way it is structured. It was hard to understand the program at first, and that meant about six weeks from starting on an idea until we finally had a successful approved launch to our project. However, once we got into it, it was a joy to learn from the more experienced mentors in HPC that were introduced to us by CAREERS at RPI.
One lesson learned was the steady reporting of progress on a firm schedule and sharing our results with other researchers, while we felt pressured some weeks, it was always a great experience once we were in the meeting, with either the bi-weekly meeting with our Mentors, or the monthly meeting with our region. This collegial feeling was a great environment to be free to ask questions and get detailed responses. One time during the monthly meeting, another student-facilitator reached out to my student-facilitator and they exchanged Python help to one another for machine learning that was a big boost to my student-facilitator, Gio.
Overall, the project was a highly ambitious one, on the edge of the latest innovations in quantum computing and machine learning, and it was pushing the boundaries of what was possible. However, because of the very high quality of collaborators, we were able to make a significant, if incremental, step forward into that mind-bending space of quantum machine learning. We succeeded in getting our research accepted for publication in the Future of Information and Communications Conference (FICC) 2022, and we presented our research in Super Computing Conference 2021 in the Birds Of A Feather session with CAREERS, and we presented our research in two posters at the Machine Intelligence Day 2021 at Pace University's Seidenberg School of Computer Science and Information Systems.