Julia: Difference between revisions

From David's Wiki
No edit summary
No edit summary
Line 101: Line 101:
would not recommend it for creating highly intricate or interactive experiences at the moment.<br>
would not recommend it for creating highly intricate or interactive experiences at the moment.<br>
MeshCat.jl is built using [https://github.com/JuliaGizmos/WebIO.jl WebIO.jl].<br>
MeshCat.jl is built using [https://github.com/JuliaGizmos/WebIO.jl WebIO.jl].<br>
===Machine Learning===
{{Main|Machine Learning in Julia}}




Line 107: Line 110:
[https://github.com/rdeits/MeshCat.jl Link]<br>
[https://github.com/rdeits/MeshCat.jl Link]<br>
A wrapper for graphics visualizations around three.js. This opens in a web browser but can be used for making helium apps.
A wrapper for graphics visualizations around three.js. This opens in a web browser but can be used for making helium apps.
See [[Graphics in Julia]] for more details on how to use this.
See [[Graphics in Julia]] for more details on how to use MeshCat.jl.
 
===Flux===
[https://fluxml.ai Link]<br>
Flux is the primary machine learning tool for Julia.
It includes convenience features including NN layers, activation functions, and an automatic differentiation system.
See [[Machine Learning in Julia]] for more details on how to use Flux.

Revision as of 20:10, 12 September 2019


Installation

Juno

Juno is an IDE for Julia. It consists of a set of packages added to Atom. Most items can be accessed from the Atom control palette Ctrl + Shift + P
Shortcuts:

  • Ctrl + Enter Evaluate current selection
  • Shift + Enter Evaluate current section and jump to next.

Basic Usage

Package Management

Guide Initializing a new project

cd project_folder
julia
] activate ./
# Add your packages

Loading an existing project

cd project_folder
julia
using Pkg;
Pkg.activate("./");
Pkg.instantiate();

String Interpolation

"Variable x is $x, y is $y, and x+y is $(x+y)"

Arrays

[Full details]

// Make an 1d array of Float64.
// Equivalent to [0.0, 0.0, 0.0]
myArr = zeros(4);

// Make a 1d array of Float64 left to uninitialized values
myArr = Array{Float64,1}(undef, 3)

// Convert an iterable to an array with collect
myArr = collect(1:3)

// Reference copy
a = [1, 2, 3];
b = [4, 5, 6];
b = a

// Shallow copy
a = [1, 2, 3];
b = [4, 5, 6];
// elementwise equals operator
b .= a
// or b[:] = a
// or b = copy(a) (this will discard the original b)

// Basic functionality
// Call functions elementwise by adding . after the function name
a = [1, 5, 99];
a = clamp.(a, 1, 11);
// a is now [1, 5, 11]
// Equivalent to clamp!(a, 1, 11)

// Julia also has map, foldl, foldr, reduce

Higher order functions

Julia supports high-order functions.

ts = ((a, b) -> (c, d) -> a + b + c + d)(1,2);

The call with arguments (1,2) returns a function.
Then ts is equivalent to

ts = (c, d) -> 1 + 2 + c + d;

You can also use the full function(a,b) syntax.

Animation Loop

You can use Timer(callback, delay, interval). This is similar to SetInterval in JavaScript.
End the loop with close(animate).

animate = Timer(function(t)
    println("Animating")
end, 0; interval=1/60)


Graphics

You can use MeshCat.jl to create visualizations to view in a web browser.
These visualizations are powered by WebGL using three.js.
Note that MeshCat.jl only exposes a small subset of three.js's features so I would not recommend it for creating highly intricate or interactive experiences at the moment.
MeshCat.jl is built using WebIO.jl.

Machine Learning


Useful Packages

MeshCat.jl

Link
A wrapper for graphics visualizations around three.js. This opens in a web browser but can be used for making helium apps. See Graphics in Julia for more details on how to use MeshCat.jl.

Flux

Link
Flux is the primary machine learning tool for Julia. It includes convenience features including NN layers, activation functions, and an automatic differentiation system. See Machine Learning in Julia for more details on how to use Flux.