I agree that this is a good place for this.

I think you can use one of the integrate functions to choose were to call oberver events (?)

That 40% figure is from Pealat and Duffy 2011.

ADE is very easy for nonlinear. Whether it is a universal panacea for all NL problems is another open question.

HMOL (odeint is VMOL)

I have the book but have not implemented it. The method is a bit esoteric, even more so than VMOL.

/// brainstorm idea; use ADE for time discretisation in combination with Meyer's HMOL in S direction. Why not?

Statistics: Posted by Cuchulainn — Today, 5:06 pm

]]>

Statistics: Posted by Cuchulainn — Today, 4:42 pm

]]>

I play around with odeint atm. Just a quick question at this point: Obersvers are not available in the implementation of the operator. This is where we calculate the derivative. And this is also the place to implement any where condition. Isnt it? The Obersver only gives us the solution once the system is integrated. But maybe i am wrong (just starting with odeint)

Things become interesting when we solve real nonlinear PDEs like UVM. What's the best methods? It is documented that ADE is 40% faster than the fastest Crank Nicolson in the West for this problem (2014).[/quote]

There is an Article from Prof. Gunter Meyer. It can be found here. http://people.math.gatech.edu/~meyer/ He does BSB with a time discrete mol. He promisses that the gamma will become smooth unlike CN or space discretization.

Statistics: Posted by berndL — Today, 1:16 pm

]]>

Statistics: Posted by Cuchulainn — September 19th, 2017, 10:45 am

]]>

Like I said, my impression that the v boundary was truncated. A remedy is to write the system of equations and then you see what v0 etc. But that is NDA!

Something like this and then it is clear.

http://ac.els-cdn.com/S0898122111002811 ... bc7e3d7e8c

v0??

stop using v0 wrongly. You're still doing it.

Statistics: Posted by outrun — September 19th, 2017, 8:52 am

]]>

About the paramters: when you say "these are the Heston parameters", then those will refer to the Heston model (i.e. the SDE parmeters) and not specific numerical methods (like grid parameters, time steps, MC samples, quadrature points,..whatever numerical method ..) . That's why it wasn't confusing.

However when you have a hammer, you see nails everywhere!

Statistics: Posted by outrun — September 19th, 2017, 8:50 am

]]>

Something like this and then it is clear.

http://ac.els-cdn.com/S0898122111002811 ... bc7e3d7e8c

Statistics: Posted by Cuchulainn — September 19th, 2017, 8:04 am

]]>

]]>

]]>

It has been explained three times now? Just read back, Billy's posts or my posts.

Statistics: Posted by outrun — September 19th, 2017, 7:40 am

]]>

]]>

I am sure he can explain. Billy, please define and put this to bed.

Statistics: Posted by Cuchulainn — September 19th, 2017, 7:34 am

]]>

]]>

outrun wrote:This thread is very strange.. Maybe you think v0 is the grid lower bound? It's not, it's the initial (latent) vol level at t=0.

You seem to be contradicting yourself.

What are you taking about. Why is this so confusing for you?

Statistics: Posted by outrun — September 19th, 2017, 7:32 am

]]>

This thread is very strange.. Maybe you think v0 is the grid lower bound? It's not, it's the initial (latent) vol level at t=0.

You seem to be contradicting yourself.

This is what Billy wrote

Statistics: Posted by Cuchulainn — September 19th, 2017, 7:30 am

]]>