julia A Fast Dynamic Language for Technical Computing Viral B. Shah Jeff Bezanson, Stefan Karpinski, Alan Edelman, and many others! Prepared for Fifth Elephant July 13, 2013

Tuesday, 10 September 13

Why do we need one more?

Tuesday, 10 September 13

Some noteworthy features ‣ Open source with an MIT licensed core ‣ Easy installation - Just download a precompiled binary and run ‣ Dynamically typed with fast user-defined types ‣ Multiple dispatch with a sophisticated parametric type system ‣ JIT compiler - no need to vectorize for performance ‣ Co-routines ‣ Distributed memory parallelism ‣ Effortlessly call C, Fortran, and Python libraries ‣ Metaprogramming with Lisp-like macros ‣ Unicode support Tuesday, 10 September 13

A simulated stock market julia> plothist(randn(100000), 100)

Tuesday, 10 September 13

julia> plot(cumsum(randn(10000)))

Let’s compute π: Buffon needle problem function buffon(m) hit = 0 for l = 1:m mp = rand() phi = (rand() * pi) - pi / 2 xrechts = mp + cos(phi)/2 xlinks = mp - cos(phi)/2 if xrechts >= 1 || xlinks <= 0 hit += 1 end end miss = m - hit piapprox = m / hit * 2 end

Tuesday, 10 September 13

Let’s compute π in parallel

function buffon_par(m) hit = @parallel (+) for l = 1:m mp = rand() phi = (rand() * pi) - pi / 2 xrechts = mp + cos(phi)/2 xlinks = mp - cos(phi)/2 (xrechts>=1||xlinks<=0) ? 1 : 0 end miss = m - hit piapprox = m / hit * 2 end

Tuesday, 10 September 13

Familiar syntax for Matlab / Octave users function randmatstat(t; n=10) v = zeros(t) w = zeros(t) for i = 1:t a = randn(n,n) b = randn(n,n) c = randn(n,n) d = randn(n,n) P = [a b c d] Q = [a b; c d] v[i] = trace((P'*P)^4) w[i] = trace((Q'*Q)^4) end std(v)/mean(v), std(w)/mean(w) end

Tuesday, 10 September 13

Keyword arguments Familiar array syntax

Common matrix operations Common statistics Last expression is return value

Yes, you can also write low-level code! Pass by reference function qsort!(a,lo,hi) i, j = lo, hi Functions ending in ! modify the while i < hi pivot = a[(lo+hi)>>>1] while i <= j while a[i] < pivot; i = i+1; end while a[j] > pivot; j = j-1; end if i <= j a[i], a[j] = a[j], a[i] Swap elements i, j = i+1, j-1 end end if lo < j; qsort!(a,lo,j); end Recursion lo, j = i, hi end return a end Tuesday, 10 September 13

inputs

Call C, Fortran, Python libraries julia> ccall(:clock, Int32, ()) 2292761 julia> ccall(:getenv, Ptr{Uint8int8}, (Ptr{Uint8},), "SHELL") Ptr{Uint8} @0x00007fff5fbffc45 julia> bytestring(ans) "/bin/bash"

julia> using PyCall # Installed with Pkg.add(“PyCall”) julia> @pyimport math julia> math.sin(math.pi / 4) - sin(pi / 4) 0.0 julia> julia> julia> julia> Tuesday, 10 September 13

@pyimport pylab x = linspace(0,2*pi,1000); y = sin(3*x + 4*cos(2*x)); pylab.plot(x, y; color="red", linewidth=2.0, linestyle="--") pylab.show()

Micro-benchmarks (log-scale) 10000

1000

100

10

1

Fortran

Julia

Python

Matlab

Octave

R

JavaScript

execution time relative to C++ (lower is better) Benchmarks: fib, parse_int, quicksort, mandel, pi_sum, rand_mat_stat, and rand_mat_mul Tuesday, 10 September 13

Let’s look at some real data julia> Pkg.add("DataFrames") Installing DataFrames: v0.3.6 julia> using DataFrames julia> vl = readtable("2013_BS_VL.csv", allowquotes=false) julia> size(vl) (1797590,12) julia> colnames(vl) julia> describe(vl)

Tuesday, 10 September 13

What does this data look like? julia> colnames(vl) 12-element Union(ASCIIString,UTF8String) Array: "AC" "ACNAME" "PS" "PSNAME" "PSADDR" "PSPART" "VoterID" "Name" "FatherHusband" "House" "Age" "Gender"

Tuesday, 10 September 13

Data is never clean julia> f = [ try int(vl["Age"][i]); catch -1; end for i in 1:size(vl,1) ] julia> vl = vl[f.!=-1, :] julia> vl[:Age] = PooledDataArray(int(vl[:Age])) julia> vl[:ACNAME] = PooledDataArray(int(vl[:ACNAME])) julia> vl[:PSNAME] = PooledDataArray(vl[:PSNAME])

julia> by(vl, :ACNAME, nrow) julia> by(vl, :ACNAME, x->mean(x[:Age])) julia> by(vl, :ACNAME, x->sum(DataArray(x[:Age] .<= 40)))

Tuesday, 10 September 13

Draw your own insights julia> describe(by(vl, :PSNAME, nrow)) PSNAME Length: 1613 Type : Pooled UTF8String NAs : 0 x1 Min 1st Qu. Median Mean 3rd Qu. Max

250.0 772.0 981.0 1113.8326100433974 1276.0 10416.0

julia> plot(1:1613, sort(pspop[:x1]))

Tuesday, 10 September 13

A great community 100+ contributors, 1000+ mailing list subscribers, 175+ packages AWS, ArgParse, BSplines, Benchmark, BinDeps, BioSeq, BloomFilters, Cairo, Calculus, Calendar, Cartesian, Catalan, ChainedVectors, ChemicalKinetics, Clang, Clp, ClusterManagers, Clustering, Codecs, CoinMP, Color, Compose, ContinuedFractions, Cpp, Cubature, Curl, DICOM, DWARF, DataFrames, DataStructures, Datetime, Debug, DecisionTree, Devectorize, DictUtils, DictViews, DimensionalityReduction, DiscreteFactor, Distance, Distributions, DualNumbers, ELF, Elliptic, Example, ExpressionUtils, FITSIO, FactCheck, FastaIO, FastaRead, FileFind, FunctionalCollections, FunctionalUtils, GLFW, GLM, GLPK, GLPKMathProgInterface, GLUT, GSL, GZip, Gadfly, Gaston, GeoIP, GeometricMCMC, GetC, GoogleCharts, Graphs, Grid, Gtk, Gurobi, HDF5, HDFS, HTTP, HTTPClient, Hadamard, HttpCommon, HttpParser, HttpServer, HypothesisTests, ICU, ImageView, Images, ImmutableArrays, IniFile, Iterators, Ito, JSON, JudyDicts, JuliaWebRepl, KLDivergence, LIBSVM, Languages, LazySequences, LibCURL, LibExpat, LinProgGLPK, Loss, MAT, MATLAB, MCMC, MDCT, MLBase, MNIST, MarketTechnicals, MathProg, MathProgBase, Meddle, Memoize, Meshes, Metis, MixedModels, Monads, Mongo, Mongrel2, Morsel, Mustache, NHST, NIfTI, NLopt, Named, NetCDF, NumericExtensions, NumericFunctors, ODBC, ODE, OpenGL, OpenSSL, Optim, Options, PLX, PTools, PatternDispatch, Phylo, Phylogenetics, Polynomial, Profile, ProgressMeter, ProjectTemplate, PyCall, PyPlot, PySide, Quandl, QuickCheck, RDatasets, REPL, RNGTest, RPMmd, RandomMatrices, Readline, Regression, Resampling, Rif, Rmath, RobustStats, Roots, SDE, SDL, SVM, SemidefiniteProgramming, SimJulia, SimpleMCMC, Sims, Sodium, Soundex, Sqlite, Stats, StrPack, Sundials, SymPy, TOML, Terminals, TextAnalysis, TextWrap, TimeModels, TimeSeries, Tk, TopicModels, TradingInstrument, Trie, URLParse, UTF16, Units, ValueDispatch, WAV, WebSockets, Winston, YAML, ZMQ, Zlib, kNN Tuesday, 10 September 13

A Fast Dynamic Language for Technical Computing - GitHub

Jul 13, 2013 - JIT compiler - no need to vectorize for performance. ‣ Co-routines. ‣ Distributed memory ... Effortlessly call C, Fortran, and Python libraries.

319KB Sizes 1 Downloads 94 Views

Recommend Documents

A Generic Language for Dynamic Adaptation
extensible dynamically, and is not generic since it is applicable only for Comet. .... In this paper we present two examples for integrating services, one is a mon-.

IARC Technical Paper - GitHub
May 28, 2017 - The teams big break came after reaching out to Dr. Maruthi Akella. ... tracks the Roombas as well as keeps its global position using data taken .... studier/emner/matnat/ifi/INF4300/h09/undervisningsmateriale/hough09.pdf.

Sample Language for Reporting and Confidentially ... - GitHub
misconduct policy.1 Schools must also consult applicable state laws (such as .... order to provide a safe, non-discriminatory environment for all students. ... the College encourages victims to talk to someone, the College provides an online [or.

Technical Drawing - GitHub
Jan 19, 2017 - SD CARD DETAIL. (OMEGA2+ ONLY). D. C. B. A. A. B. C. D. 1. 2. 3. 4. 5. 6. 7. 8. 8. 7. 6. 5. 4. 3. 2. 1. OMEGA2. TITLE. DIMENSION IN MM.

The Ruby Programming Language - GitHub
You'll find a guide to the structure and organization of this book in Chapter 1. ..... Determine US generation name based on birth year ...... curly braces: "360 degrees=#{2*Math::PI} radians" # "360 degrees=6.28318530717959 radians" ...... of comput

VDM-10 Language Manual - GitHub
Overture Technical Report Series. No. TR-001. May 2017 ... Document history. Month. Year Version Version of Overture.exe Comment. April ... Contents. 1 Introduction. 1. 1.1 The VDM Specification Language (VDM-SL) . ..... 13.3.1 Classes .

Computing Dynamic Optimal Mechanisms When ...
Sep 25, 2011 - University of Wisconsin, Madison. Yuichiro ..... If the environment is convex, B maps convex sets into convex ...... Princeton University Press.

Scientific Computing for Biologists Hands-On Exercises ... - GitHub
Oct 25, 2011 - Discriminant Analysis in R. The function ... data = iris, prior = c(1, 1, 1)/3) ... if we were analyzing a data set with tens or hundreds of variables.