Re: Not to incite a language war but apparently the Corona lockdown was based on 13 year old undocumented C-Code

[Niklas Holsti <niklas.holsti@tidorum.invalid>]

On 2020-05-13 0:27, Jeffrey R. Carter wrote:
> On 5/12/20 9:16 PM, Niklas Holsti wrote:
>>>
>>> It seems clear that the code that was executed to create the 
>>> predictions was not deterministic on a single core.
>>
>> The first of the two cases of single-core "non-determinism" described 
>> and bewailed in the "review" is that the time evolution of the 
>> predicted epidemic is different depending on the value of a program 
>> option -- whether the "network" is created anew, or reused -- that 
>> should not affect the result. No actual non-determinism is claimed in 
>> the review, just that the epidemic reaches large infection rates 
>> sooner for one option value than for the other. It seems that the code 
>> was using the PRN generator slightly differently in the two cases, 
>> resulting in different PRN sequences.
> 
> The single-core non-determinism that I gathered from the referenced 
> interpretation of the review, which I clearly have not studied as 
> carefully as you, was that the program gave different results on the 
> same computer for identical inputs: same initial conditions and same RNG 
> seed(s). Perhaps I misunderstood, but that was what I got from it, and 
> the only part that seemed troubling. This is different from the kind of 
> non-determinism of simulations that are intended to be run many times 
> with different RNG seeds. Possibly that would be caused by accessing 
> uninitialized variables.

The "second analysis" at 
https://lockdownsceptics.org/second-analysis-of-fergusons-model/ refers 
to some GitHub change-log entries that do correct uninitialized-variable 
errors, so there may have been some real non-determinism.

The second analysis then tries to evaluate the effect of the various 
errors and their corrections by running the newest version and the 
"initial import" version with the same data, and claims "radically 
altered" results.

In fact, the results compared, which are the number of infected but 
recovered people on day 128 of the simulation, are some 57 million for 
the initial code and 42 million for the current code. But note that the 
reviewer compares only one run of each program, so I would say that this 
change can well be purely statistical and result from small random 
(PRN-induced) differences in the speed at which the infections take off 
in the two runs.

Even if the two runs use the same initial RNG seed, the changes probably 
make the sequence of RNG calls, and the uses to which the RNG results 
are put, different in the two program versions, which means that they 
effectively use different PRN sequences.

I don't intend to spend any more time on this storm in a teapot.

-- 
Niklas Holsti
niklas holsti tidorum fi
       .      @       .