{"id":4619,"date":"2012-10-07T11:24:43","date_gmt":"2012-10-07T10:24:43","guid":{"rendered":"http:\/\/www.walkingrandomly.com\/?p=4619"},"modified":"2012-10-28T15:34:11","modified_gmt":"2012-10-28T14:34:11","slug":"european-option-pricing-in-julia-and-matlab","status":"publish","type":"post","link":"https:\/\/walkingrandomly.com\/?p=4619","title":{"rendered":"European Option Pricing in Julia and MATLAB"},"content":{"rendered":"

I felt like playing with Julia<\/a> and MATLAB<\/a> this Sunday morning.\u00a0 I found some code that prices European Options<\/a> in MATLAB using Monte Carlo simulations over at computeraidedfinance.com<\/a> and thought that I’d port this over to Julia.\u00a0 Here’s the original MATLAB code<\/p>\n

function V = bench_CPU_European(numPaths)\r\n%Simple European\r\nsteps = 250;\r\nr = (0.05);\r\nsigma = (0.4);\r\nT = (1);\r\ndt = T\/(steps);\r\nK = (100);\r\n\r\nS = 100 * ones(numPaths,1);\r\n\r\nfor i=1:steps\r\n rnd = randn(numPaths,1);\r\n S = S .* exp((r-0.5*sigma.^2)*dt + sigma*sqrt(dt)*rnd);\r\nend\r\nV = mean( exp(-r*T)*max(K-S,0) )<\/pre>\n
I ran this a couple of times to see what results I should be getting and how long it would take for 1 million paths:<\/p>\n
tic;bench_CPU_European(1000000);toc\r\nV =\r\n   13.1596\r\nElapsed time is 6.035635 seconds.\r\n>> tic;bench_CPU_European(1000000);toc\r\nV =\r\n   13.1258\r\nElapsed time is 5.924104 seconds.\r\n>> tic;bench_CPU_European(1000000);toc\r\nV =\r\n   13.1479\r\nElapsed time is 5.936475 seconds.<\/pre>\n
The result varies because this is a stochastic process but we can see that it should be around 13.1 or so and takes around 6 seconds on my laptop. Since it’s Sunday morning, I am feeling lazy and have no intention of considering if this code is optimal or not right now. I’m just going to copy and paste it into a julia file and hack at the syntax until it becomes valid Julia code. The following seems to work<\/p>\n
function bench_CPU_European(numPaths)\r\n\r\nsteps = 250\r\nr = 0.05\r\nsigma = .4;\r\nT = 1;\r\ndt = T\/(steps)\r\nK = 100;\r\n\r\nS = 100 * ones(numPaths,1);\r\n\r\nfor i=1:steps\r\n rnd = randn(numPaths,1)\r\n S = S .* exp((r-0.5*sigma.^2)*dt + sigma*sqrt(dt)*rnd)\r\nend\r\nV = mean( exp(-r*T)*max(K-S,0) )\r\nend<\/pre>\n
I ran this on Julia and got the following<\/p>\n
julia> tic();bench_CPU_European(1000000);toc()\r\nelapsed time: 36.259000062942505 seconds\r\n36.259000062942505\r\n\r\njulia> bench_CPU_European(1000000)\r\n13.114855104505445<\/pre>\n
The Julia code appears to be valid, it gives the correct result of 13.1 ish but at 36.25 seconds is around 6 times slower<\/strong> than the MATLAB version.\u00a0 The dog needs walking so I’m going to think about this another time but comments are welcome.<\/p>\n
Update (9pm 7th October 2012):<\/strong>\u00a0\u00a0 I’ve just tried this Julia code on the Linux partition of the same laptop and 1 million paths took 14 seconds or so:<\/p>\n
tic();bench_CPU_European(1000000);toc()\r\nelapsed time: 14.146281957626343 seconds<\/pre>\n
I built this version of Julia from source and so it’s at the current bleeding edge (version 0.0.0+98589672.r65a1 Commit 65a1f3dedc (2012-10-07 06:40:18). The code is still slower than the MATLAB version but better than the older Windows build<\/p>\n
Update: 13th October 2012<\/strong><\/p>\n
Over on the Julia mailing list<\/a>, someone posted a faster version of this simulation in Julia<\/p>\n
function bench_eu(numPaths)\r\n    steps = 250\r\n    r = 0.05\r\n    sigma = .4;\r\n    T = 1;\r\n    dt = T\/(steps)\r\n    K = 100;\r\n\r\n    S = 100 * ones(numPaths,1);\r\n\r\n    t1 = (r-0.5*sigma.^2)*dt\r\n    t2 = sigma*sqrt(dt)\r\n    for i=1:steps\r\n        for j=1:numPaths\r\n            S[j] .*= exp(t1 + t2*randn())\r\n        end\r\n    end\r\n\r\n    V = mean( exp(-r*T)*max(K-S,0) )\r\nend<\/pre>\n
On the Linux partition of my test machine, this got through 1000000 paths in 8.53 seconds, very close to the MATLAB speed:<\/p>\n
julia> tic();bench_eu(1000000);toc()\r\nelapsed time: 8.534484148025513 seconds<\/pre>\n
It seems that, when using Julia, one needs to unlearn everything you’ve ever learned about vectorisation in MATLAB.<\/p>\n
Update: 28th October 2012<\/strong><\/p>\n
Members of the Julia team have been improving the performance of the randn() function used in the above code (see here<\/a> and here<\/a> for details).\u00a0 Using the de-vectorised code above, execution time for 1 million paths in Julia is now down to 7.2 seconds on my machine on Linux.\u00a0 Still slower than the MATLAB 2012a implementation but it’s getting there.\u00a0 This was using Julia version \u00a00.0.0+100403134.r0999 Commit 099936aec6 (2012-10-28 05:24:40)<\/p>\n
tic();bench_eu(1000000);toc()\r\nelapsed time: 7.223690032958984 seconds\r\n7.223690032958984<\/pre>\n