<t>QuoteOriginally posted by: outrunThe Sobol sequences fill the hypercube with low discrepancy, which is different than keeping the mean to zero.A solution would be to duplicate the results of iii & multiply them with -1, genenerating two random sample vectors out of each Sobol draw (antithetic...

Large Numbers

Perhaps you find this paper by Glasserman interesting.Malliavin Greeks Without Malliavin Calculus

Not that I'm an FX expert but could you elaborate a bit?

<t>QuoteIrrespective of which norm is used (pointwise errors are useful) it is so that the schemes in the original post are disappointing. Maybe I've coded it incorrectly...Cuch, may I ask some questions?Have you found a discretization for Heston which didn't disappoint you?What Heston parameters di...

Latest work are pre-fascicles 0b 0c and 1a which can be found here.

<t>For the record, I just noticed ACE also supports Linux's epoll, albeit with the still experimental Dev_Poll_Reactor. EDIT: I just checked with the developers. Here's what they say:QuoteACE_Dev_Poll_Reactor is fully functional and supported with Linuxepoll. It has not been as well exercised with /...

<t>QuoteIn the employ's perspective, will they value the BSD socket programming skills more or they like to see the familarity with certain package like ACE. I'd say a basic understanding of BSD sockets is always necessary. This was so in the RPC days and I don't see why it should have changed. When...

<t>QuoteOriginally posted by: dibbleQuoteOriginally posted by: QuantBitWe're using ACE and just started using the less bulky libevent.I just had a quick look at libevent. When you say it is less bulky, do you mean that it has a lot less bells and whistles, but it does have the bells and whistles tha...

We're using ACE and just started using the less bulky libevent.

<t>ACE has a big user community which includes industry and finance:Here are three finance examples from the previous link:"ACE is used for a global limiting system at Credit Suisse, Zurich, Switzerland.""Bear Stearns utilizes ACE in Risk Analysis Control System (RACS) to facilitate its server/clien...

BS diffusion is exponential. Any suggestion why this is not so for the models you're using?Vague wording alert: Jump diffusion produces fat tails but still is locally exponential, isn't it?

Monte Carlo and even better, Quasi-Monte Carlo, should also work. Especially when your x has many dimensions.

<r>We're using Finite Differences. All standard methods work fine for us. That is Explicit, Fully Implicit and Crank Nicolson. Sorry Cuch, have not yet gotten around to implement Exponential Fitting <E>;-)</E>.Problems we're having from time to time is exploding or complex Local Volatility when usin...

The number of reads per window matters. If the number of reads is very big, you want a data structure that is fast to read, like an array or vector. If the number of reads is small, a tree structure is preferable because it is updated faster.