Skip to content

bottydim/ads8-neuralnets

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 

Repository files navigation

ADS Template Repository

Note to editors: make sure you have nbstripout installed on your machine so that you don't clog the repository with Jupyter outputs!

This folder has the structure that all ADS modules are expected to have.

Generic schedule

  • 1.50 - 1.50 (0930-1100) Block 0
  • 0.25 - 1.75 (1100-1115) Coffee break
  • 1.50 - 3.25 (1115-1245) Block 1
  • 1.00 - 4.25 (1245-1345) Lunch
  • 1.50 - 5.75 (1345-1515) Block 2
  • 0.25 - 6.00 (1515-1530) Coffee break
  • 1.50 - 7.30 (1530-1700) Block 3

Note Block 0 for all modules on day 1 will be a discussion (either of code, or about the programme or some such).

Side stuff

  • Type A questions: "exact" questions (implement code that does X)
  • Type B questions: "open" questions (do something that answers a generic question)

Presentation

  • For each module starting after module 2, every team has to prepare a short presentation (10 minutes).
  • Three teams are randomly selected (tweak to have everyone present a bit).
  • These teams present one after the other.
  • 10-15 minutes questions / comments directly addressed to teams.
  • We present our comments, a standard solution, also discuss Type A questions

Leaderboard concept

They get points for every module following:

If present

  • +10 for presenting
  • +10a where a is the ratio of people present (non jury) who voted for the presentation as the best one
  • +10 if jury voted it the best

All

  • points for code cleanness (1..10)
  • points for code quality (1..10) (test A questions)
  • points for model quality (3-5-10) (test B questions, top three of model accuracy leaderboard if relevant. If ties, all people tying get the maximum score available to them.)

Team randomisation

For 28 people:

s = shuffle(15:28)
[(i, s[i]) for i in 1:14]

yields

 (1, 23)
 (2, 17)
 (3, 28)
 (4, 16)
 (5, 21)
 (6, 22)
 (7, 25)
 (8, 15)
 (9, 20)
 (10, 26)
 (11, 19)
 (12, 24)
 (13, 18)
 (14, 27)

this is an assignment for the first module. Then you just cycle up one for each module. No overlap in 9 weekends. Don't pre-announce everything, just say it's randomly drawn every time and announce it 2nd day.

About

ADS Course Neural Network Module

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages