InputValues = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
md `## Graph Plot Setting`
height =700
width =1000
maxLayer =d3.max(data.nodes, d => d.LayerNum)
maxwt = d3.max(data.links, d => Math.abs(d.weight))
maxnode = d3.max(data.nodes, d => Math.abs(d.value))
color = {
const scale = d3.scaleOrdinal(d3.schemeCategory10);
return d => scale(d.LayerNum);
}
drag = simulation => {
function dragstarted(d) {
if (!d3.event.active) simulation.alphaTarget(0.3).restart();
d.fx = d.x;
d.fy = d.y;
}
function dragged(d) {
d.fx = d3.event.x;
d.fy = d3.event.y;
}
function dragended(d) {
if (!d3.event.active) simulation.alphaTarget(0);
d.fx = null;
d.fy = null;
}
return d3.drag()
.on("start", dragstarted)
.on("drag", dragged)
.on("end", dragended);
}
d3 = require("d3@5")
import {select} from "@jashkenas/inputs"
md `## Graph Generation`
NN = d3.json(model)
InputNames = ['phi_0', 'theta_0', 'x_99', 'y_99', 'phi_99', 'theta_99', 'xdot_0', 'ydot_0', 'phidot_0', 'thetadot_0']
function ComputeLayerValues2(input){
let nodes = [];
let links = [];
let connected = [];
for (let i=0; i<NN['weights'][0].length; i++){
let node = {'id': i, 'LayerNum': 0, 'NodeNum': InputNames[i], 'value': input[i], 'valueScaled': input[i]}
nodes.push(node)
}
let currentLayer = [].concat(input)
for (let l=0; l<NN['weights'].length; l++){
const weight = NN['weights'][l]
let bias = [].concat(NN['bias'][l])
let nextLayer =[].concat(bias)
for (let i=0; i<weight.length; i++){
for (let j=0; j<weight[i].length; j++){
nextLayer[j] += weight[i][j]*currentLayer[i]
let link = {'source': l*NN['WidthLimit']+i, 'target': (l+1)*NN['WidthLimit']+j, 'weight': weight[i][j]}
if (Math.abs(link.weight) > 0.001 || NN.WidthLimit<=100){
links.push(link)
if (!connected.includes(link.source)){
connected.push(link.source)
}
if (!connected.includes(link.target)){
connected.push(link.target)
}
}
}
}
for (let k=0; k<nextLayer.length; k++){
let v = nextLayer[k]
if (l !== NN['weights'].length-1){
v = Math.tanh(nextLayer[k]) // apply activation function
let node = {'id': (l+1)*NN['WidthLimit']+k, 'LayerNum': l+1, 'NodeNum': k, 'value': v, 'valueScaled': v}
nodes.push(node)
nextLayer[k] = v
}
if (l === NN['weights'].length-1){
let node = {'id': (l+1)*NN['WidthLimit']+k, 'LayerNum': l+1, 'NodeNum': OutputNames[k], 'value': v, 'valueScaled': Math.tanh(v)}
nodes.push(node)
}
}
currentLayer = [].concat(nextLayer)
}
let nodesConnect = []
for (let n=0; n<nodes.length; n++){
if (connected.includes(nodes[n].id)){
nodesConnect.push(nodes[n])
}
}
return {"nodes": nodesConnect, "links": links};
}
function RecomputeLayerValues2(input){
let nodes = [];
let links = [];
let connected = [];
for (let i=0; i<NN['weights'][0].length; i++){
let node = {'id': i, 'LayerNum': 0, 'NodeNum': InputNames[i], 'value': input[i], 'valueScaled': input[i]}
document.getElementById("node_"+i).setAttribute("r",Math.abs(2*input[i])*10+3);
document.getElementById("node_"+i).setAttribute("fill" , input[i]>0? "#FF8C00":"#00CED1");
document.getElementById("title_"+i).textContent ='layer: 0, node: '+ InputNames[i] + ', value: '+ input[i];
nodes.push(node)
}
let currentLayer = [].concat(input)
for (let l=0; l<NN['weights'].length; l++){
const weight = NN['weights'][l]
let bias = [].concat(NN['bias'][l])
let nextLayer =[].concat(bias)
for (let i=0; i<weight.length; i++){
for (let j=0; j<weight[i].length; j++){
nextLayer[j] += weight[i][j]*currentLayer[i]
let link = {'source': l*NN['WidthLimit']+i, 'target': (l+1)*NN['WidthLimit']+j, 'weight': weight[i][j]}
if (Math.abs(link.weight) > 0.001 || NN.WidthLimit<=100){
links.push(link)
if (!connected.includes(link.source)){
connected.push(link.source)
}
if (!connected.includes(link.target)){
connected.push(link.target)
}
}
}
}
for (let k=0; k<nextLayer.length; k++){
let v = nextLayer[k]
if (l !== NN['weights'].length-1){
v = Math.tanh(nextLayer[k]) // apply activation function
let node = {'id': (l+1)*NN['WidthLimit']+k, 'LayerNum': l+1, 'NodeNum': k, 'value': v, 'valueScaled': v}
var temp = (l+1)*NN['WidthLimit']+k
document.getElementById("node_"+temp.toString()).setAttribute("r" , Math.abs(Math.tanh(nextLayer[k]))*10+3);
document.getElementById("node_"+temp.toString()).setAttribute("fill" , v>0? "#FF8C00":"#00CED1");
document.getElementById("title_"+temp.toString()).textContent ="layer: "+ l+1 +", node: " + k + ', value: '+ Math.tanh(nextLayer[k]);
nodes.push(node)
nextLayer[k] = v
}
if (l === NN['weights'].length-1){
let node = {'id': (l+1)*NN['WidthLimit']+k, 'LayerNum': l+1, 'NodeNum': OutputNames[k], 'value': v, 'valueScaled': Math.tanh(v)}
var temp = (l+1)*NN['WidthLimit']+k
document.getElementById("node_"+temp.toString()).setAttribute("r" , Math.abs(Math.tanh(nextLayer[k]))*10+3);
document.getElementById("node_"+temp.toString()).setAttribute("fill" , v>0? "#FF8C00":"#00CED1");
document.getElementById("title_"+temp.toString()).textContent ="layer: "+ l+1 +", node: " + OutputNames[k] + ', value: '+ Math.tanh(nextLayer[k]);
nodes.push(node)
}
}
currentLayer = [].concat(nextLayer)
}
let nodesConnect = []
for (let n=0; n<nodes.length; n++){
if (connected.includes(nodes[n].id)){
nodesConnect.push(nodes[n])
}
}
return {"nodes": nodesConnect, "links": links};
}
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.