How to tune hyper parameters to train an agent based on proximal policy optimization in RL?

Damsara Udan Jayarathne @Damsara-Jayarathne

31 August 2022 0 5K Report

Hello everyone,

I am working on a problem in which a point mass is trying to catch another point mass. The dynamics are correct and implemented in MATLAB RL uswing a PPO agent. I have normalised the reward and the observations. The reward is 2*exp(-0.005*(e_x^2+e_z^2)) where e_x and e_z refer to the error in x and z directions.

I am a little bit confused about tuning the hyper parameters. Currently I am using

"ExperienceHorizon",100,...

"ClipFactor",0.1,...

"EntropyLossWeight",0.01,...

"MiniBatchSize",75,...

"NumEpoch",5,...

"AdvantageEstimateMethod","gae",...

"GAEFactor",0.95,...

Total training time per episode is 10 seconds (100 steps). The agent does not seem to train even after 20000 iteration.

I would be grateful if someone could point me in the right direction. Thanks in advance.

Badges
Science topic

Why should we choose deep learning to machine learning? Why is CNN better for image classification? Can I get Typical Answer?

What Characteristics makes CNN work better?

03 March 2021 1,458 4 View

What are some of the research gaps in machine learning and artificial intelligence in africa?

i would to know some of the research gaps in the artificial intelligence field in most african countries.

03 March 2021 6,145 3 View

Anyone plz suggest relevant topic for research in image processing+Deep learning.?

I have selected brain tumor images ...but now found that already lots of research done n this topic.

03 March 2021 5,774 3 View

How to increase a model accuracy ?

dear community, my model is based feature extraction from non stationary signals using discrete Wavelet Transform and then using statistical features then machine learning classifiers in order to...

03 March 2021 6,994 5 View

Can you share me how teacher education is managed in your respective national educational system?

I feel that the practice in teacher education in my country is below the expected performance level due to very poor management system. Hope I will learn something from your experiences.

02 March 2021 1,516 4 View

Can NFL theorem be valid in infinite search space in ML?

NFL theorem is valid for algorithms training in fixed training set. However, the general characteristic of algorithms in expanded or open dataset has not been proved yet. Could you show your...

01 March 2021 1,189 3 View

Is there any research regarding how training and development is going to change in the era of Fourth Industrial Revolution?

As the organizations getting prepared for the fourth industrial revolution, to enhance the skills of their workforce what type of changes are necessary and going to occur in the training and...

01 March 2021 3,199 3 View

Timeout error!! Processor in the loop (PIL) using STM32F4 Matlab 2018a?

Dear Researchers I am trying to perform a PIL simulation using STM32F4 Discovery board and comunication serial USB TO TTL. During simulation I receive the following timeout error: An error...

01 March 2021 2,327 1 View

Which group formation theory do you recommend for projects that aim to bring together entrepreneurs for training for regional development?

I would like to have ideas about which theories of group formation are suitable for entrepreneurial training projects, with the aim of creating bonds, building trust, and collaboration for...

28 February 2021 4,223 2 View

What is the difference between L1 and L2 regularization in machine learning?

L1 and L2 regularization

28 February 2021 4,187 3 View