[TWIM Notes] Dec 8 2020

Message boards : News : [TWIM Notes] Dec 8 2020
Message board moderation

To post messages, you must log in.

AuthorMessage
pianoman [MLC@Home Admin]
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Jun 20
Posts: 446
Credit: 14,122,724
RAC: 36,468
Message 922 - Posted: 9 Dec 2020, 2:15:19 UTC

This Week in MLC@Home
Notes for Dec 8 2020
A weekly summary of news and notes for MLC@Home

Summary
Lots of analysis of the results, and paper writing, and fighting with LaTeX this week. The good news is, as posted on Twitter, Dataset 3 shows similar networks clustering in weight-space the same as dataset 1 and 2 do. This is a nice result and currently being written up as part of a "dataset paper" we're preparing for when we release the datasets to the public. We're currently training our own neural network classifier to classify which DS3 networks were trained with which DS3 data based solely on their weights, and we have high hopes for good results there given the clustering in the linked graph.

It was a a relatively quiet week on the forums, but big leaps in WU progress! We're less than 900 networks away from completing the EightBitModified networks for DS2, less than 9000 networks away from completing DS3 milestone 2, and less than 80 ParityModified networks away from completing DS2-500. Thanks again to
all our volunteers.

News:

  • DS1 and DS2 continue towards completion. We're down to less than a thousand left for EightBitModified, leaving only ParityMachine/Modified left for DS1 and DS2.
  • DS3 progress is approaching milestone 2, 1000x100.. 100000 DS3 networks trained. That will definitely help with our ongoing analysis. It's nice to see the scoreboard at the bottom of https://www.mlcathome.org/ turning from yellow to green!
  • We've created the DS1-100, DS1-500, DS1-1000, DS2-100, DS3-100, DS3-500 datasets internally, and are cleaning them up and analyzing the results. We'd like to get at least DS2-500, DS2-1000, and DS3-1000 done ASAP for our analysis (DS2-1000 might be a stretch).
  • Once Dataset 3 reaches milestone 2 (100x10000), we'll likely pause DS3 to focus only on completing Parity WUs for finishing off DS1 and DS2.
  • We haven't talked about DS4 in a while, but work progresses on the updated client to handle CNNs. Expect that to be released during or just after the holidays.



Project status snapshot:
(note these numbers are approximations)

Tasks
Tasks ready to send 22937
Tasks in progress 21142
Users
With credit 1406
Registered in past 24 hours 63
Hosts
With recent credit 2314
Registered in past 24 hours 25
Current GigaFLOPS 44334.91

Dataset 1 and 2 progress:

SingleDirectMachine      10002/10004
EightBitMachine          10001/10006
SingleInvertMachine      10001/10003
SimpleXORMachine         10000/10002
ParityMachine             1179/10005

ParityModified             428/10005
EightBitModified          9155/10006
SimpleXORModified        10005/10005
SingleDirectModified     10004/10004
SingleInvertModified     10002/10002 

Dataset 3 progress:
Overall (so far): 91044/101488
Milestone 1, 100x100:  10000/10000
Milestone 2, 100x1000: 91044/100000
Milestone 3: 100x10000: 91044/1000000


Last week's TWIM Notes: Nov 30 2020

Thanks again to all our volunteers!

-- The MLC@Home Admins(s)
Homepage: https://www.mlcathome.org/
Twitter: @MLCHome2
ID: 922 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Haced

Send message
Joined: 25 Dec 20
Posts: 2
Credit: 0
RAC: 0
Message 971 - Posted: 25 Dec 2020, 10:09:12 UTC - in response to Message 922.  

Can i get a project
ID: 971 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : News : [TWIM Notes] Dec 8 2020

©2021 MLC@Home Team
A project of the Cognition, Robotics, and Learning (CORAL) Lab at the University of Maryland, Baltimore County (UMBC)