Public
Edited
Jan 16, 2023
11 forks
40 stars
params = ({
mu_k: 4.0, sigma_k: 1.0, w_k: 0.022,
mu_g: 0.6, sigma_g: 0.15, c_rep: 1.0,
dt: 0.1
})
point_n = 200
function init(points) {
for (let i=0; i<point_n; ++i) {
points[i*2] = (Math.random()-0.5)*12;
points[i*2+1] = (Math.random()-0.5)*12;
}
return points;
}
points = init(new Float32Array(point_n*2));
fields = ({
R_val: new Float32Array(point_n),
U_val: new Float32Array(point_n),
R_grad: new Float32Array(point_n*2),
U_grad: new Float32Array(point_n*2),
})
function add_xy(a, i, x, y, c) {
a[i*2] += x*c; a[i*2+1] += y*c;
}
function compute_fields() {
const {R_val, U_val, R_grad, U_grad} = fields;
const {c_rep, mu_k, sigma_k, w_k} = params;
// account for the own field of each particle
R_val.fill(repulsion_f(0.0, c_rep)[0]);
U_val.fill(peak_f(0.0, mu_k, sigma_k, w_k)[0]);
R_grad.fill(0); U_grad.fill(0);
for (let i=0; i<point_n-1; ++i)
for (let j=i+1; j<point_n; ++j) {
let rx = points[i*2] - points[j*2];
let ry = points[i*2+1] - points[j*2+1];
const r = Math.sqrt(rx*rx + ry*ry) + 1e-20;
rx /= r; ry /= r; // ∇r = [rx, ry]
if (r < 1.0) {
// ∇R = R'(r) ∇r
const [R, dR] = repulsion_f(r, c_rep);
add_xy(R_grad, i, rx, ry, dR);
add_xy(R_grad, j, rx, ry, -dR);
R_val[i] += R; R_val[j] += R;
}
// ∇K = K'(r) ∇r
const [K, dK] = peak_f(r, mu_k, sigma_k, w_k);
add_xy(U_grad, i, rx, ry, dK);
add_xy(U_grad, j, rx, ry, -dK);
U_val[i] += K; U_val[j] += K;
}
}
function repulsion_f(x, c_rep) {
const t = Math.max(1.0-x, 0.0);
return [0.5*c_rep*t*t, -c_rep*t];
}
function fast_exp(x) {
let t = 1.0 + x/32.0;
t *= t; t *= t; t *= t; t *= t; t *= t; // t **= 32
return t;
}
function peak_f(x, mu, sigma, w=1.0) {
const t = (x-mu)/sigma;
const y = w / fast_exp(t*t);
return [y, -2.0*t*y/sigma];
}
function step() {
const {R_val, U_val, R_grad, U_grad} = fields;
const {mu_g, sigma_g, dt} = params;
compute_fields();
let total_E = 0.0;
for (let i=0; i<point_n; ++i) {
const [G, dG] = peak_f(U_val[i], mu_g, sigma_g);
// [vx, vy] = -∇E = G'(U)∇U - ∇R
const vx = dG*U_grad[i*2] - R_grad[i*2];
const vy = dG*U_grad[i*2+1] - R_grad[i*2+1];
add_xy(points, i, vx, vy, dt);
total_E += R_val[i] - G;
}
return total_E / point_n;
}
function animate(ctx, world_width=25.0) {
for (let i=0; i<steps_per_frame; ++i) step();
const {width, height} = ctx.canvas;
ctx.resetTransform();
ctx.clearRect(0, 0, width, height);
ctx.translate(width/2, height/2);
const s = width/world_width;
ctx.scale(s, s);
ctx.lineWidth = 0.1;
for (let i=0; i<point_n; ++i) {
ctx.beginPath();
const x=points[i*2], y=points[i*2+1];
const r = params.c_rep / (fields.R_val[i]*5.0);
ctx.arc(x, y, r, 0.0, Math.PI*2);
ctx.stroke();
}
}
One platform to build and deploy the best data apps
Experiment and prototype by building visualizations in live JavaScript notebooks. Collaborate with your team and decide which concepts to build out.
Use Observable Framework to build data apps locally. Use data loaders to build in any language or library, including Python, SQL, and R.
Seamlessly deploy to Observable. Test before you ship, use automatic deploy-on-commit, and ensure your projects are always up-to-date.