Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Quantum Reinforcement Learning related projects #35

Open
mehilagarwal opened this issue Aug 23, 2021 · 12 comments
Open

Quantum Reinforcement Learning related projects #35

mehilagarwal opened this issue Aug 23, 2021 · 12 comments

Comments

@mehilagarwal
Copy link

Description

I do not have a specific project in mind. It would be great if there is a mentor interested in this topic. I would want to work on research on Quantum Reinforcement Learning algorithms or implementing existing research using Qiskit and if that exists then benchmarking those implementations.

Mentor/s

Looking for mentors

Type of participant

<What are the profiles of the ideal participants for this idea?>

Number of participants

1-3

Deliverable

Either a prototype; paper if possible. Or, Qiskit code/library for Quantum RL algorithms or benchmarking results

@HuangJunye
Copy link
Collaborator

@amitracal @olgaatgh @vkumar19 @dalevras @deepquantum88 Can you please comment in the issue so that I can assign you?

@deepquantum88
Copy link

deepquantum88 commented Sep 22, 2021 via email

@vkumar19
Copy link

Commenting here

@da66
Copy link

da66 commented Sep 22, 2021

Commenting...

@amitracal
Copy link

Commenting

@olgaatgh
Copy link

Commenting here

@amitracal
Copy link

#35 Quantum Reinforcement Learning.pdf

@HuangJunye HuangJunye added checkpoint1: submitted and removed status: matched The project is matched and will not take any more mentees labels Oct 6, 2021
@mehilagarwal
Copy link
Author

The first phase of the project is to include reinforcement learning (RL) as a classical optimization algorithm to support the quantum approximate optimization algorithm (QAOA).
At each predefined step, we update the observation_space (QAOA parameter space) based on the information stored in the active_space (allowed actions and rewords). Then the expectation value is calculated on the quantum backend (or simulator) following the problem definition. If the result of the quantum circuit execution is leading the optimization process in the desirable direction, we “issue” the reward as a part of the “reinforcement” and update observation_space accordingly. Otherwise, we penalize the unfavorable choice of parameters. The algorithm runs cyclically for the given number of iterations.
In the given architecture the QAOA environment is RL formulation agnostic, meaning it can accept value function or policy function optimization including the Actor-Critic method(A2C), Proximal Policy Optimization algorithm(PPO), and Deep Q Network(DQN).
We currently testing this approach using a well-known max-cut problem and planning to look at more advanced problems while refining the current QAOA-RL implementation
The pdf attached represents the crux of our project in the form of a flowchart.
Flowchart.pdf

I am working with @amitracal @olgaatgh @da66 @vkumar19! Thank you @deepquantum88 for your guidance!

@HuangJunye
Copy link
Collaborator

The mentees can you please upload the final presentation here?

@amitracal
Copy link

QRL#35_Dec9 (1).pptx

@amitracal
Copy link

@amitracal
Copy link

@HuangJunye both presentations for the two checkpoints are uploaded now
@olgaatgh @deepquantum88 @da66 @vkumar19 @mehilagarwal FYI

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment