MichalisPanayides / PhD-meetings

This is a repository to keep track of the weekly meetings for the purposes of my PhD

0 stars 0 forks source link

2022-04-26 meeting overview #88

Open MichalisPanayides opened 2 years ago

MichalisPanayides commented 2 years ago

What has been done:

Review for JOSS ✔️
Registered for CORS/INFORS and EURO2022 ✔️
Booked everything for Canada trip ✔️
Julia Scientific Programming course ⌚
THE paper:
- Addressed all comments ✔️
- PDF with annotated changes ✔️
- Resubmitted ✔️
Greggynog Talk ✔️
More data/plots on RL algorithm (still running on siren) ⌚
Slight improvement on source code 👉 Getting the transition matrix is a lot faster now (see comments)⌚

To be discussed:

[x] Show the annotated .pdf (mainly because it took me a whole day to generate and I want to show it off 😄)
[x] Reinforcement Learning - look at the new plots
[x] Greggynog talk + what to change for CORS/INFORMS & EURO
[x] Thesis
- Similar reviewing procedure to tha paper?
- Am I to include any source code in the actual thesis?

To be done:

Annual Review 😢
Thesis
Finish up all talks to be given
- Greggynog
- CORS/INFORS
- Netherlands talk with Paul
- EURO2022
More data on RL
Other things I want to do:
- Direct formula for blocking time
- Closed form formula for steady states probabilities

MichalisPanayides commented 2 years ago

MichalisPanayides commented 2 years ago

Hackathon 29/04/22 - TODO's:

[x] Merge improve-transition-matrix-implementation branch
Reinforcement Learning:
- [x] Look at how the matrices change (DO NOT BE LAZY AND LOOK AT THE MATRICES)
- [ ] Plot moving averages of the service rate graph
- [ ] Merge utility plot and service rate plot into one (utilities bold, rates dashed)
- [x] Closely look (zoom in) at plot when e=0.1

Non-hackathon TODOs:

[ ] Try different starting policies
[ ] Try different set of parameters
[ ] Plot the mean service rate instead of the mean of all service rates
- Somehow get the probability distribution of all states and times it with each server's rate
[ ] Update the RL algorithm
- Use probability distribution of the states and use that to pick the new state to update
- Probabilistically based on that probability distribution

Williamsem commented 2 years ago

Hackathon 29/04/22 - TODO's:

[x] Decide Research Objectives
- How these have been met so far
[x] Make Gantt Chart
- Decide what is left to do
[ ] Make list of thesis chapters
[ ] Complete Research Proposal
[ ] Excel tool interface
[x] Book Vancouver hotel
[ ] Book Rochester hotel
[x] Email KESS
[x] Email Maths finance