SERVING THE QUANTITATIVE FINANCE COMMUNITY

Cuchulainn
Posts: 50458
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

outrun wrote:
Would be cool to see how Cuch's framework and tricks would preform relative to Billy7 working solver which seems to work great.

What would be the timeline, soon -or- someday but not anytime soon?

Soon, before end April 2017. It's a hard deadline.  In fact, several of us have coded it up but not exactly for Heston and I want to put in the new stuff for Heston.
http://www.datasimfinancial.com

“The two most important days in your life are the day you are born and the day you find out why.”

Cuchulainn
Posts: 50458
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

Often, too low accuracy in the objective function will hinder convergence. If your objective function has accuracy that is too low, the optimizer can get confused, and run for 100 steps, never converging. With higher accuracy, it might run 15 steps (each 'step' involving say 20 evaluations of the objective function) and converge.

So, you are saying that the particular solver used is sensitive to noise and/or initial guess?
Maybe a solver that is less sensitive?
http://www.datasimfinancial.com

“The two most important days in your life are the day you are born and the day you find out why.”

Alan
Posts: 8857
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

I assume you meant "optimizer" not "solver". Yeah, in Mathematica anyway, it is a trade-off.

FindMinimum is very sensitive to noise, and if there is too much, it will just cycle. NMinimize (using the Differential Evolution method) is less sensitive to noise, but typically is much slower.
Last edited by Alan on February 16th, 2017, 3:23 pm

outrun
Posts: 2420
Joined: April 29th, 2016, 1:40 pm

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

Cuchulainn wrote:
outrun wrote:
Would be cool to see how Cuch's framework and tricks would preform relative to Billy7 working solver which seems to work great.

What would be the timeline, soon -or- someday but not anytime soon?

Soon, before end April 2017. It's a hard deadline.  In fact, several of us have coded it up but not exactly for Heston and I want to put in the new stuff for Heston.

Nice.
You keep saying Heston, but you mean Garch (the topic of Alan's goal), right?

Cuchulainn
Posts: 50458
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

outrun wrote:
Cuchulainn wrote:
outrun wrote:
Would be cool to see how Cuch's framework and tricks would preform relative to Billy7 working solver which seems to work great.

What would be the timeline, soon -or- someday but not anytime soon?

Soon, before end April 2017. It's a hard deadline.  In fact, several of us have coded it up but not exactly for Heston and I want to put in the new stuff for Heston.

Nice.
You keep saying Heston, but you mean Garch (the topic of Alan's goal), right?

I mean indeed Heston PDE and its FD approximation which is one the themes here (it's  project in itself in any case).. If I have understood Billy7 properly, GARCH can be done based on it?(?)
http://www.datasimfinancial.com

“The two most important days in your life are the day you are born and the day you find out why.”

outrun
Posts: 2420
Joined: April 29th, 2016, 1:40 pm

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

Cuchulainn wrote:
outrun wrote:
Cuchulainn wrote:
Soon, before end April 2017. It's a hard deadline.  In fact, several of us have coded it up but not exactly for Heston and I want to put in the new stuff for Heston.

Nice.
You keep saying Heston, but you mean Garch (the topic of Alan's goal), right?

I mean indeed Heston PDE and its FD approximation which is one the themes here (it's  project in itself in any case).. If I have understood Billy7 properly, GARCH can be done based on it?(?)

Yes is was a theme here, but Alan's challenge is specifically about calibration the GARCH parameters. Maybe we should move that challenge to a new thread?

Cuchulainn
Posts: 50458
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

There seems to be several sub-projects going on here. Maybe someone can scope them so that they can be worked on independently?
http://www.datasimfinancial.com

“The two most important days in your life are the day you are born and the day you find out why.”

outrun
Posts: 2420
Joined: April 29th, 2016, 1:40 pm

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

Cuchulainn wrote:
There seems to be several sub-projects going on here. Maybe someone can scope them so that they can be worked on independently?

Yes. One active one is Alan's challenge, with 4 sub challenges, $15.000 price money each afaik. Billy7 Posts: 176 Joined: March 30th, 2016, 2:12 pm ### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE. Cuchulainn wrote: outrun wrote: Cuchulainn wrote: Soon, before end April 2017. It's a hard deadline. In fact, several of us have coded it up but not exactly for Heston and I want to put in the new stuff for Heston. Nice. You keep saying Heston, but you mean Garch (the topic of Alan's goal), right? I mean indeed Heston PDE and its FD approximation which is one the themes here (it's project in itself in any case).. If I have understood Billy7 properly, GARCH can be done based on it?(?) Yes, that was my experience, though I haven't been able to test more, maybe today. As can readily be seen by comparing the Heston and GARCH pricing PDEs, the only thing different is the power of v in the diffusion and cross derivative terms. So the mechanics should be the same, but obviously stability characteristics would change. Other details change as well, but overall going from one to the other shouldn't be much of a problem. Last edited by Billy7 on February 17th, 2017, 12:58 pm Cuchulainn Posts: 50458 Joined: July 16th, 2004, 7:38 am Location: Amsterdam Contact: ### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE. outrun wrote: Cuchulainn wrote: There seems to be several sub-projects going on here. Maybe someone can scope them so that they can be worked on independently? Yes. One active one is Alan's challenge, with 4 sub challenges,$15.000 price money each afaik.

Do we get 20% up front for initial layout costs?
http://www.datasimfinancial.com

“The two most important days in your life are the day you are born and the day you find out why.”

Cuchulainn
Posts: 50458
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

outrun wrote:
Cuchulainn wrote:
There seems to be several sub-projects going on here. Maybe someone can scope them so that they can be worked on independently?

Yes. One active one is Alan's challenge, with 4 sub challenges, $15.000 price money each afaik. Do we get 20% up front for initial layout costs? http://www.datasimfinancial.com “The two most important days in your life are the day you are born and the day you find out why.” Billy7 Posts: 176 Joined: March 30th, 2016, 2:12 pm ### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE. outrun wrote: Cuchulainn wrote: There seems to be several sub-projects going on here. Maybe someone can scope them so that they can be worked on independently? Yes. One active one is Alan's challenge, with 4 sub challenges,$15.000 price money each afaik.

Damn, enough fooling around then, I'll finish all 4 today!

Cuchulainn
Posts: 50458
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

A micro optimiser regarding Godonov Double Sweep vs Thomas (ADI)

We have stress-tested these solvers when applied to the heat equation that we discuss in section 13.3 using mesh sizes of 100,000 in space and time. Both methods give accurate results (even though the double sweep tend to be more accurate). The processing times in seconds were in the ranges [490, 588] for double sweep and [730, 780] for the Thomas algorithm.
http://www.datasimfinancial.com

“The two most important days in your life are the day you are born and the day you find out why.”

Cuchulainn
Posts: 50458
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE.

Billy7 wrote:
outrun wrote:
Cuchulainn wrote:
There seems to be several sub-projects going on here. Maybe someone can scope them so that they can be worked on independently?

Yes. One active one is Alan's challenge, with 4 sub challenges, $15.000 price money each afaik. Damn, enough fooling around then, I'll finish all 4 today! Is that Eastern Standard Time or Rocky Mountain Time? Don't tell me., Daylight Saving Time https://www.youtube.com/watch?v=l5Kdc0LLSW8 Or as Fred Brooks say Q How does a s/w project get behind schedule? A. One day at a time. http://www.datasimfinancial.com “The two most important days in your life are the day you are born and the day you find out why.” Billy7 Posts: 176 Joined: March 30th, 2016, 2:12 pm ### Re: PDE methods for optimal quantizers of stochastic processes ? An example with Heston PDE. Cuchulainn wrote: Billy7 wrote: outrun wrote: Yes. One active one is Alan's challenge, with 4 sub challenges,$15.000 price money each afaik.

Damn, enough fooling around then, I'll finish all 4 today!

Is that Eastern Standard Time or Rocky Mountain Time? Don't tell me., Daylight Saving Time

Or as Fred Brooks say
Q How does a s/w project get behind schedule?
A. One day at a time.

Indeed, I guess it's gonna have to wait till tomorrow then:)