Posts by Aurum

1) Questions and Answers : Issue Discussion : Too many WUs (Message 268)
Posted 28 Jul 2020 by Aurum
Post:
One option is to emulate what WCG did on their Device Profiles page.
Now all my MLC rigs are overloaded. I'll try 0.1/0.1 and see if that helps.
2) Questions and Answers : Issue Discussion : Adaptive replication (Message 264)
Posted 27 Jul 2020 by Aurum
Post:
How long before I graduate to Lone Ranger status of being considered a highly reliable host???
https://boinc.berkeley.edu/trac/wiki/AdaptiveReplication
3) Message boards : Science : News Article (Message 263)
Posted 27 Jul 2020 by Aurum
Post:
https://phys.org/news/2020-07-machine-reveals-recipe-artificial-proteins.amp
ā€¯Though artificial intelligence revealed the design rules, Ranganathan and his collaborators still don't fully understand why the models work. Next they will work to understand just how the models came to this conclusion. "There is much more work to be done," he said.
...the studies in proteins might even help teach us how the deep neural networks behind modern machine learning actually work."
The most important point in this article is, "The model shows that just conservation at amino acid positions and correlations in the evolution of pairs of amino acids are sufficient to predict new artificial sequences that would have the properties of the protein family."
By conserved they mean it does not change. In nature this means that if highly conserved amino acids are changed then the function and/or structure changes so much that evolution will eliminate it. What they fail to tell us is what percentage of the chorismate mutase proteins are conserved in their experiment. Then they have a second rule "correlations in the evolution of pairs of amino acids" that further reduces the number of possibilities. This work is not as profound as the author implies it is. I'd enjoy reading the "An evolution-based model for designing chorismate mutase enzymes, Science (2020)" article but I'm not going to pay $15 for it.
4) Questions and Answers : Issue Discussion : Too many WUs (Message 261)
Posted 27 Jul 2020 by Aurum
Post:
After MLC becomes the only project running on a computer it seems to behave better.
Maybe the machine learned ;-)
5) Message boards : Cafe : Badges (Message 259)
Posted 27 Jul 2020 by Aurum
Post:
I like a brainy style badge but any decent image looks awful at 30 pixels. Use a larger badge size like GPUGrid & LHC do.



6) Questions and Answers : Issue Discussion : Too many WUs (Message 258)
Posted 27 Jul 2020 by Aurum
Post:
My BOINC queue is set to 1.0/0.1 days but MLC sends me far more WUs than I can possibly finish in a day or before the deadline in 2 days.
This causes a couple of problems. A small one is if one wants to transition from say WCG to MLC the excess MLC WUs cause all MLC WUs to "Run as High Priority." This means BOINC does not timeslice back to WCG to finish them off. Have to babysit until WCG WUs are cleared out.
Bigger issue is that you need fast turn around time for your WUs. Sending too many guarantees you slow yourself down since many WUs will come back as unstarted.
7) Message boards : News : [TWIM Notes] Jul 25, 2020 (Message 257)
Posted 27 Jul 2020 by Aurum
Post:
Forum posts are fine with me. But I have never had an admin ask if he was giving us too much news before.
I will have to recover from it.
Ditto. More likely to get a deaf ear :-)
8) Questions and Answers : Issue Discussion : Validation errors. (Message 256)
Posted 27 Jul 2020 by Aurum
Post:
If this project allows some artificial intelligence machine to discover a cure for COVID-19, then the validity errors won't matter.
Is this a goal of this project? First I've seen this but it's only my second day here and I'm still reading.




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