Contact CDT and clear out the port issue

• Food (e.g. pizza for lunch)
• After party with one free beer (or not)

• Posters at Universities (PW, UW, etc.)
• Show up on facebook groups (students, DL, ML, coding related)
• Ask student associations for a help

Check possibilities to easy to start hyperparameters search for participants (like Optima)

Initial R&D

This is an update referenced to #12

Start with testing following

• Test: Create a batch of 10 sub-accounts.
• Test: Remove/Delete the 10 sub-accounts.
• Test: Create a batch of sub-accounts with a service deployed on each of them(ex: 1 VM each)
• Test: Remove/Delete the 10 sub-accounts with VM's

Research

• Can we add say 4 people to the same sub-account, So that all team members will get on the same sub-account?
• If above = true, Can we run say 1 server that the 4 users can access at the same time ( _basically this will be more cost-efficient + it adds the factor that either @raznem or @jacekplocharczyk mentioned, that the team will be able to affect how they utilize the resources and also make it fair play regardless of a team have 2 or 5 members, they will have the same type of resource)

Handling data around accounts
Through dll-web-app we will be able to pull data in various forms. We need to establish the format of the data used for cloud-formation

• Establish format of data, User by user or users grouped by team and then inform @Mindgames so an export function for Firestore can be developed that will pull & transform data as requested

References

https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/Welcome.html
https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-resource-sagemaker-notebookinstance.html

When we will have food and after party we could set up some time window for their 20-30 min. RL introduction.

Each participant should be familiar with coding on at least basic level.

• Local leader board for Warsaw DLL
• Local leader board for each DLL location
• Global leader board
• How to manage people who are changing teams? Teams leader board (for stable teams) and individual leader board (for participants)?

https://www.kaggle.com/c/connectx

Introductory RL lecture in separate room for 20-30 mins

DLL Global #1

04.04.2020

This time we will have a standard deep learning challenge.

General Info:

Challenge: Predict COVID-19 spreading across specific countries and worldwide
Criterion: Root mean squared logarithmic error (RMSLE)

Additional challenge: Find meaningful insights about COVID-19 (an open question)

Test set: see general TODOs

Dataset info/TODOs:

1. We could use the data from the Kaggle competition (Alternative: ECDC data)
   I think that the license check is required.
2. We could use the data from the World Bank about economical indicators for each country (some helping materials)
   Manual data extraction could be required - we can use this code or it's also possible to download .csv files directly from their page
3. We could find a dataset with travel restrictions, airports shutdowns, etc. to combine it with the rest of the data.
   Example of such data: https://www.kaggle.com/paultimothymooney/covid19-containment-and-mitigation-measures
4. We could use climate data for each country.

General TODOs:

1. Describe evaluation metrics for the main challenge
   We could use column-wise root mean squared logarithmic error like in the Kaggle competition
   We will use RMSLE for evaluation.
2. Define evaluation datasets
   We could prepare two test sets - one from last week (public used only for evaluation if the submission is working properly, it could overlap with the trainset) and the other based on the week after hackathon with public leaderboard updated daily. @raznem @Mindgames what do you think?
3. Should we predict infected or deaths?
   It is a delicate matter so we should be careful here.
4. Describe evaluation metrics for the additional challenge
   @Mindgames please let us know when you'll find someone (or some company/organization)

We will update this post when we find out more.

• What GitHub pages framework do we want to use?
• Decide on content
• Deploy

• Make notebook with a fancy visualized environment
• Simple tasks for newbies in this notebook
• Work on some hints for basic tasks

AWS cloud formation will allow us to run DLL in scale while staying consistent and allow participants to access other services & functions on the AWS platform

• Understand the ecosystem around CF and how it works. @mathias
• Decide on setup to implement
• Confirm function
• Decide on how participents will recive their login & server information