Paper Elsevier Journal MLWA 2022

[detlefarend]

Detlef:

• [x] Title
• [x] Abstract
• [x] 1 Introduction
• [x] 2 Fundamentals of MLPro
• [x] 3 MLPro-RL - The Sub-Framework for Reinforcement Learning
• [x] 6 Conclusion and Future Work

Steve:

• [x] 4 MLPro-GT - The Sub-Framework for Game Theory
• [x] 5.a Simulation of a Bulk Goods Plant

Rizky:

• [x] 5.b Control of a UR Robot Arm

Review:

• [x] Review 1 (all authors)
• [x] Review 2 (Prof. Schwung)

Submission:

• [x] Paper submitted (Detlef)

Journal: https://www.journals.elsevier.com/machine-learning-with-applications

[detlefarend]

Hi colleagues, I just added the abstract. Could you please take a look whether it's good enough? If ok so far I could create the related ResearchGate project (after Prof. Schwung agreed to this as well). Thx in advance!

[steveyuwono]

Hi Detlef. It looks good and straightforward. However, I found some wordy sentences as a reader. May I modify some sentences? As far as I'm concerned, the abstract in IEEE should be written in a single paragraph.

[detlefarend]

Hi Steve, yes please. Thank you!

[steveyuwono]

I've modified some sentences. You can check the updated abstract in our overleaf.

[detlefarend]

I've modified some sentences. You can check the updated abstract in our overleaf.

Hi Steve, thanks for your suggestions. That helped me to reflect on the text and to improve some of the wording. The thing is that I have sent the origin text to Prof. Schwung as well and he agreed to it w/o changes. So I didn't want to change too much afterwards.

[steveyuwono]

Okay great! If he agreed, then we don’t need to change too much

[detlefarend]

Hi guys, good news: the submission deadline is November 30th.

[detlefarend]

Hi colleagues,

I finished my parts and started reviewing everything. Here are some annotations on my part:

Section 3: MLPro-RL Would be nice if you could review it. Is it correct, logical, understandable? Is something important missing?

Section 4: MLPro-GT @steveyuwono: have the explanations in Sec.3 an impact to this section?

Section 5: Sample Apps Here are some hints from my perspective that hopefully help to unify and focus the sample sections: I think we can roughly break down each sample into three parts and the following aspects are rather meaned as a kind of checklist to ensure that we don't forget something:

a) Description of the scenario

• Environment/Game Board (technical name/rtd url as footnote)
• Agent/Player structure, Policy algorithms (SB3 algo with wrapper, ...)
• Scenario and related Howto if public (technical name/rtd url as footnote). If not public we should annotate the reason.
• Training setup: especially the termination criteria. Here I please you to try the new evaluation strategy with stagnation detection.
• Nice appetizing picture(s)

b) Training results

• Evaluation based plot
• Highscore, Number of training cycles/episodes

c) Summary What did we demonstrate?

Subsection 5a: BGLP The policy algo is private and as such not part of MLPro. Shall we describe this shortly?

Subsection 5b: UR5 Robot @rizkydiprasetya: Shall we show both robot pictures (sim/real) side by side? Are you fine with repeating the training with evaluation/stagnation or is there any difficulty (except runtime)?

Section 6: Conclusion Please review and let me know whether we can optimize something.

Thx!!

[rizkydiprasetya]

ok. For now only runtime. I will try to run with evaluation/stagnation on environment robothtm.