Gradient Descent (iterative) algorithm … by rodrigo.siqueira

function u(t) {

c.width=1920 // clear the canvas
for(i=0;i<9;i++)
x.fillRect(400+i*100+S(t)*300,400,50,200) // draw 50x200 rects

} // 122/140

Please log in (or register) to post a dweet (copy-paste the code somewhere safe to save it meanwhile)


    u(t) is called 60 times per second.
    t: Elapsed time in seconds.
    S: Shorthand for Math.sin.
    C: Shorthand for Math.cos.
    T: Shorthand for Math.tan.
    R: Function that generates rgba-strings, usage ex.: R(255, 255, 255, 0.5)
    c: A 1920x1080 canvas.
    x: A 2D context for that canvas.

show FPS hide FPS share fullscreen

rodrigo.siqueira

function u(t) {

η=.1
ℵ=1/t
for(X=-3;X<3;X+=.01){
y=X**3-X*2
ℵ=ℵ-η*(ℵ**3-2*ℵ)
x.fillRect(940+X*400,-y*100+500,5,5)
}
x.fillRect(910+ℵ*400,470,60,60)
throw(ℵ)

Aug 21 2019 1:45 PM UTC

}// 140/140

Please log in (or register) to post as a new dweet (copy-paste code somewhere safe to save it meanwhile).

Gradient Descent (iterative) algorithm used to calculate sqrt(2). The big dots are the two roots -sqrt(2) and +sqrt(2) of equation Y=X**3-X*2. The Gradient Descent is an optimization algorithm for finding a local minimum of a function near a starting point, taking successive steps in the direction of the negative of the gradient. x[n+1]=x[n]−η(∇f)xn where η is the "Learning Rate". Gradient Descent is widely used for training Machine Learning models. #neural #ai

u/rodrigo.siqueira

Simple version that calculates sqrt(2). If you change initial "x" to -1, it will converge to -sqrt(2), that is the other root of the equation Y=X**3-X*2. Code: x = 1; learning_rate = .1; for(i=0;i<50;i++) { x = x - learning_rate * (x**3 - 2*x) } throw(x);


      u(t) is called 60 times per second.
      t: elapsed time in seconds.
      c: A 1920x1080 canvas.
      x: A 2D context for that canvas.
      S: Math.sin
      C: Math.cos
      T: Math.tan
      R: Generates rgba-strings, ex.: R(255, 255, 255, 0.5)

dwitter.net