Latest Coronavirus - Yikes

#56,126
#56,126
  • Like
Reactions: TrumpedUpVol, 508mikey and Orange_Vol1321
#56,128
#56,128
  • Like
Reactions: Franklin Pierce and BigOrangeD
#56,130
#56,130
Viruses don't discriminate based on political affiliation, so I thought it was interesting that 77% of COVID deaths have occurred in Blue states, even though there are fewer 22 vs 28 and less population. There was also a recent study that found nearly every case found in the southern states originated in NY.
 
  • Like
Reactions: NurseGoodVol, allvol123, davethevol and 1 other person
#56,131
#56,131
TennTradition said:
If you google around you can find the article the software CEO wrote - and this my comments around software. He harps on their coding for most of the article.

But he makes the case that these coding issues leave the mode to produce different results when different computers run the code. What’s not clear to me is if that means Ferguson got an incorrect result - or if the model was producing correct results on their lab’s computers.

It isn’t clear at all to me from this that there was an error in their results at the time - other than just the inputs/assumptions. For example, given what we know now a 2.2 Mm death toll in the US doesn’t make sense unless you have higher death rates due to hospital overburden. But, this is just a software guy saying it was really bad code. I need him to tell me - were there coding errors that led Ferguson to get an incorrect result (or just model inaccuracies due to assumptions)
Click to expand...
Thanks TT. Without reading the entire article, I can only speculate, but it sounds like this CEO is taking shots at the now-infamous Ferguson model (accurate shots or not, I can't judge at this point) in order to gain some street cred as a software developer. But in my understanding, "bad code" does not necessarily mean you don't get the same result(s) from a given model and set of input data compared to "good code" but that "good code" may be more efficient in its use of computing resources.
 
  • Like
Reactions: BigOrangeD
#56,133
#56,133
My mom has a guy doing tile work for her, he went to a conference in LV back in January for his wife’s work. He said there were a lot of people there from China and most of them had masks on. A couple days after they got home the wife could not stop coughing. In a couple days he had the same thing, then the guy working for him got it. The wife went to the doctor and was negative for flu. My guess is that’s three more people that will test positive on an antibody screen, and I think we’re going to slowly push back the timeline on when this thing actually took off and came to the US.
 
  • Like
Reactions: NurseGoodVol, allvol123 and davethevol
#56,134
#56,134
RavinDave said:
Thanks TT. Without reading the entire article, I can only speculate, but it sounds like this CEO is taking shots at the now-infamous Ferguson model (accurate shots or not, I can't judge at this point) in order to gain some street cred as a software developer. But in my understanding, "bad code" does not necessarily mean you don't get the same result(s) from a given model and set of input data compared to "good code" but that "good code" may be more efficient in its use of computing resources.
Click to expand...

Normally I would agree with you.

Somehow, this guy was arguing the model due to its architecture wasn’t deterministic. I need to understand that. On the same computer, do I get a different answer every run? Was that intentional?

Or is it a compiling issue on different machines? If so, how did it compile on Ferguson’s lab’s computers.
 
#56,135
#56,135
508mikey said:
idiots
AP_20124461653670.jpg
Click to expand...
turd worlders
 
  • Like
Reactions: 508mikey
#56,139
#56,139
W.TN.Orange Blood said:
Here you go America. Go outside and enjoy life but do this to stay safe ..... Is this the future for us?

View attachment 280233
Click to expand...

I see two people laying on the grass outside of the circles , do they have to pay a fine or just get shamed into an empty circle? Also I was thinking wouldn’t it be better for the city to put collars around the necks of the ones in the circles like if they step outside they get a shock ? They could rent them out for like an hour at a time and make money off them . Kinda like car parking .
 
  • Like
Reactions: Carl Pickens and W.TN.Orange Blood
#56,140
#56,140
0nelilreb said:
I see two people laying on the grass outside of the circles , do they have to pay a fine or just get shamed into an empty circle? Also I was thinking wouldn’t it be better for the city to put collars around the necks of the ones in the circles like if they step outside they get a shock ? They could rent them out for like an hour at a time and make money off them . Kinda like car parking .
Click to expand...

I also noticed after looking again ...... there's nobody wearing masks?
 
#56,142
#56,142
TennTradition said:
Normally I would agree with you.

Somehow, this guy was arguing the model due to its architecture wasn’t deterministic. I need to understand that. On the same computer, do I get a different answer every run? Was that intentional?

Or is it a compiling issue on different machines? If so, how did it compile on Ferguson’s lab’s computers.
Click to expand...
Unless the model includes a randomizing function, shouldn't one always expect the same output from the same input on a given computer?

Software can be computer- or compiler-specific, in which case if trying to run it on a different system or compiler version than it was written for I would expect incompatibility or compiler errors i.e. failure to execute rather than deviations in results. I still don't see how the "architecture" of the model is causing it to deviate when executing in a different computing environment unless the model is intended to do that.
 
#56,143
#56,143
RavinDave said:
Unless the model includes a randomizing function, shouldn't one always expect the same output from the same input on a given computer?

Software can be computer- or compiler-specific, in which case if trying to run it on a different system or compiler version than it was written for I would expect incompatibility or compiler errors i.e. failure to execute rather than deviations in results. I still don't see how the "architecture" of the model is causing it to deviate when executing in a different computing environment unless the model is intended to do that.
Click to expand...

Agreed - that’s why I was questioning the compiler.

I feel like this was a hit piece by someone not in the field that is pissed off - so they railed against what they know: software architecture.

However, they got my attention when they started talking about the model being non-deterministic due to “wildly different” results on different machines and implied it wasn’t randomness by design.

I’d like to understand why the different results, were Ferguson’s 2.2 and 500k predictions the result of software error, and what “wildly different” means. The authors didn’t address those in detail.
 
Last edited:
#56,145
#56,145
0nelilreb said:
I see two people laying on the grass outside of the circles , do they have to pay a fine or just get shamed into an empty circle? Also I was thinking wouldn’t it be better for the city to put collars around the necks of the ones in the circles like if they step outside they get a shock ? They could rent them out for like an hour at a time and make money off them . Kinda like car parking .
Click to expand...
35DB69A5-111B-4695-9036-6347DF0FE3CF.gif
 
  • Like
Reactions: 0nelilreb
#56,149
#56,149
TennTradition said:
Most of this just seems to be attacking their code architecture - which isn’t surprising given these guys aren’t software developers. They aren’t trying to build code for retail sale so that seemed like a silly complaint.

However - their points on the model not being deterministic are concerning at face value. I really need to see some plots/analysis to understand how sensitive the code is to the computer running it.
Click to expand...
You should change your username to Mr. Peabody
 
New posts
You must log in or register to reply here.

VN Store



Back
Top