SERVING THE QUANTITATIVE FINANCE COMMUNITY

 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 4:53 pm

That's why bank have validation teams, you need to validate things before you use them.
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 5:11 pm

That's why bank have validation teams, you need to validate things before you use them.
Indeed! The question is: do the validation teams understand the downstream and long-term implications of the code, model, and it's output? That is, the code might be correct WRT the model but lead to a financial crisis.
 
User avatar
Cuchulainn
Posts: 60432
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 6:15 pm

That's why bank have validation teams, you need to validate things before you use them.
Image
On the plus, no one dies validating financial reports; MRI data is somewhat different.
Last edited by Cuchulainn on August 3rd, 2017, 7:33 pm, edited 1 time in total.
 
User avatar
Cuchulainn
Posts: 60432
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 6:21 pm

A similar but even more insidious problem happened with some high-end Xerox copier/scanners systems that used an especially aggressive adaptive image compression scheme. Sometimes the compression encoder would mis-recognize small-type numbers and replace them with other numbers.

Exactly! and that is the real issue here; how 'consolidated' gets changed and interpreted in going from CAD to CAM.
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 8:00 pm

That's why bank have validation teams, you need to validate things before you use them.
Image
On the plus, no one dies validating financial reports; MRI data is somewhat different.
Testing is not validating. It would be very easy to see the missing < as being invalid, right?
 
User avatar
Cuchulainn
Posts: 60432
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 8:10 pm

That's why bank have validation teams, you need to validate things before you use them.
Image
On the plus, no one dies validating financial reports; MRI data is somewhat different.
Testing is not validating. It would be very easy to see the missing < as being invalid, right?
Should be possible, indeed; MRI depts have a FDA guy on board..they missed the issue?
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 8:42 pm

Image
On the plus, no one dies validating financial reports; MRI data is somewhat different.
Testing is not validating. It would be very easy to see the missing < as being invalid, right?
Should be possible, indeed; MRI depts have a FDA guy on board..they missed the issue?
My brother had been making some certified medical stuff (a blood cooler) and it's *extremely* expensive to get a certificate, mainly because of all the analysis external parties have to do to make sure it works as intended. When I validate a model (not that often) it's also extremely expensive because of the huge amount of time it takes to go through all the details and have 3rd parties validate the validation process.

Lately there is a lot of talk about easy hackable medical software. That's imo also very irresponsible to neglect.
 
User avatar
Cuchulainn
Posts: 60432
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 10:11 pm

Coders in these orgs tend to write about 2 lines of code per two weeks. 
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 11:56 pm

Testing is not validating. It would be very easy to see the missing < as being invalid, right?
Should be possible, indeed; MRI depts have a FDA guy on board..they missed the issue?
My brother had been making some certified medical stuff (a blood cooler) and it's *extremely* expensive to get a certificate, mainly because of all the analysis external parties have to do to make sure it works as intended. When I validate a model (not that often) it's also extremely expensive because of the huge amount of time it takes to go through all the details and have 3rd parties validate the validation process.

Lately there is a lot of talk about easy hackable medical software. That's imo also very irresponsible to neglect.
Indeed! And about medical devices running nasty consumer-grade operating systems and having things happen such as ceasing to work during a surgical procedure because Windows Update kicked in and took over the machine.
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 3rd, 2017, 11:57 pm

Coders in these orgs tend to write about 2 lines of code per two weeks. 
Yes and they are the best damned two lines you'll ever see! (Too bad they still get things wrong like metric vs. imperial, overflows on float->int conversion, etc.)
 
User avatar
Cuchulainn
Posts: 60432
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 4th, 2017, 6:49 am

Coders in these orgs tend to write about 2 lines of code per two weeks. 
Yes and they are the best damned two lines you'll ever see!  (Too bad they still get things wrong like metric vs. imperial, overflows on float->int conversion, etc.)
LOL

I thought ppauper was the only one who dreams in Fahfahrenheit!
 
User avatar
Traden4Alpha
Posts: 23951
Joined: September 20th, 2002, 8:30 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 4th, 2017, 11:50 am

Coders in these orgs tend to write about 2 lines of code per two weeks. 
Yes and they are the best damned two lines you'll ever see!  (Too bad they still get things wrong like metric vs. imperial, overflows on float->int conversion, etc.)
LOL

I thought ppauper was the only one who dreams in Fahfahrenheit!
Indeed! I grew up in a country too arrogant to change from a obsolete system of measurement.

-----

The problem with validators is that they live and breath in the same technical & mathematical culture as the programmers. If the underlying models are wrong, the validators may not catch it because the validators are using the same models & maths, too. That is, validators can check consistency but not correctness.
 
User avatar
outrun
Posts: 4573
Joined: April 29th, 2016, 1:40 pm

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 4th, 2017, 12:29 pm

Coders in these orgs tend to write about 2 lines of code per two weeks. 
Yes and they are the best damned two lines you'll ever see!  (Too bad they still get things wrong like metric vs. imperial, overflows on float->int conversion, etc.)
In hardware it's common to do "formal verifications" but you don't see that much in software.. There is however e.g. the sel4 microkernel:
Unique about seL4 is its unprecedented degree of assurance, achieved through formal verification. Specifically, the ARM version of seL4 is the first (and still only) general-purpose OS kernel with a full code-level functional correctness proof, meaning a mathematical proof that the implementation (written in C) adheres to its specification. In short, the implementation is proved to be bug-free (see below). This also implies a number of other properties, such as freedom from buffer overflows, null pointer exceptions, use-after-free, etc.
What puzzles me is that Intel at some point had a Pentium line that had a floating point computations bug. This means it wasn't formally verified. Any background on this?
 
User avatar
Cuchulainn
Posts: 60432
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 4th, 2017, 12:37 pm

Yes and they are the best damned two lines you'll ever see!  (Too bad they still get things wrong like metric vs. imperial, overflows on float->int conversion, etc.)
LOL

I thought ppauper was the only one who dreams in Fahfahrenheit!
Indeed!  I grew up in a country too arrogant to change from a obsolete system of measurement.

-----

The problem with validators is that they live and breath in the same technical & mathematical culture as the programmers.  If the underlying models are wrong, the validators may not catch it because the validators are using the same models & maths, too.  That is, validators can check consistency but not correctness.
Indeed. Validators are probably not allowed to challenge the status quo.
All models are wrong, some are useful. When models stagnate, it begins to look like a hammer and then they needs nails.
 
User avatar
Cuchulainn
Posts: 60432
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Grammar-Pattern Programmers: Is Over-Abstraction Delivering Projects Faster Or Slower?

August 4th, 2017, 2:09 pm

Question: the medical "< problem".. which of these is it?

An error is a wrong decision that was made during the development of a software product. A defect is a property of a software system that may cause that system to depart from its intended behaviour. A fault is the event of a software system departing from its intended behaviour during one of its executions.

So, when things go wrong who should you approach? 
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On