TensorFlowNotes

load tensorflow

<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@0.12.5"></script>

create tensors

// create a vector: sampleVec = tf.tensor([1,3,7,9])
// create a matrix or a higher dimensional tensor: tf.tensor([[1, 2], [3, 4]]);
// create a matrix with a flat array and a shape: tf.tensor([1, 2, 3, 4], [2, 2]);
// to enhance code readability, use: tf.scalar, tf.tensor1d, tf.tensor2d, tf.tensor3d and tf.tensor4d.

	// Define a model for linear regression.
	const model = tf.sequential();
	model.add(tf.layers.dense({units: 1, inputShape: [1]}));

	// Prepare the model for training: Specify the loss and the optimizer.
	model.compile({loss: 'meanSquaredError', optimizer: 'sgd'});

	// Generate some synthetic data for training.
	const xs = tf.tensor2d([1, 2, 3, 4], [4, 1]);
	const ys = tf.tensor2d([1, 3, 5, 7], [4, 1]);

	// Train the model using the data.
	model.fit(xs, ys, {epochs: 10}).then(() => {
	  // Use the model to do inference on a data point the model hasn't seen before:
	  // Open the browser devtools to see the output
	  model.predict(tf.tensor2d([5], [1, 1])).print();
	});

some tensor ops

square:
=======
const d = tf.tensor2d([[1.0, 2.0], [3.0, 4.0]]);
const d_squared = d.square();

d_squared.print();
// Output: [[1, 4 ],
//          [9, 16]]

add:
====
const e = tf.tensor2d([[1.0, 2.0], [3.0, 4.0]]);
const f = tf.tensor2d([[5.0, 6.0], [7.0, 8.0]]);

const e_plus_f = e.add(f);
e_plus_f.print();
// Output: [[6 , 8 ],
//          [10, 12]]

chainable of ops:
=================
const sq_sum = e.add(f).square();
sq_sum.print();
// Output: [[36 , 64 ],
//          [100, 144]]

const sq_sum = tf.square(tf.add(e, f));

model

a model is a function

two ways to create models, use ops directly to represent the work the model does.
For example:

// Define function
function predict(input) {
  // y = a * x ^ 2 + b * x + c
  return tf.tidy(() => {
    const x = tf.scalar(input);

    const ax2 = a.mul(x.square());
    const bx = b.mul(x);
    const y = ax2.add(bx).add(c);

    return y;
  });
}

// Define constants: y = 2x^2 + 4x + 8
const a = tf.scalar(2);
const b = tf.scalar(4);
const c = tf.scalar(8);

// Predict output for input of 2
const result = predict(2);
result.print() // Output: 24


the other way is to use the tf.model to construct a model out of layers, which are a popular abstraction in deep learning.

The following code constructs a tf.sequential model:

const model = tf.sequential();
model.add(
  tf.layers.simpleRNN({
    units: 20,
    recurrentInitializer: 'GlorotNormal',
    inputShape: [80, 4]
  })
);

const optimizer = tf.train.sgd(LEARNING_RATE);
model.compile({optimizer, loss: 'categoricalCrossentropy'});
model.fit({x: data, y: labels});

There are many different types of layers available in TensorFlow.js. 
A few examples include tf.layers.simpleRNN, tf.layers.gru, and tf.layers.lstm.

http://qt.gtimg.cn/q=

http://qt.gtimg.cn/q=
http://qt.gtimg.cn/r=2&q=r_hk01072

var script = document.createElement('script');
script.type = 'text/javascript';
script.src = 'https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@0.6.1';
document.head.appendChild(script);




$.getScript('https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@0.6.1');
$.getScript('http://qt.gtimg.cn/r=2&q=r_hk01072"');

v_r_hk01072

string.split(separator, limit)

jsonp colllect external file

<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.3.1/jquery.min.js">

$(document).ready(function(){
}

    s.src = "http://qt.gtimg.cn/r=2&q=r_hk01072?callback=myDisplayFunction";
    document.body.appendChild(s);
	//$("head").append(script);
	//document.getElementsByTagName('head')[0].appendChild(script);
	// or document.head.appendChild(script) in modern browsers


    $("button").click(function(){
        $.get("http://qt.gtimg.cn/r=2&q=r_hk01072", function(data, status){
		  console.log(data);
            alert("Data: " + data + "\nStatus: " + status);
        });
    });

    $("button").click(function(){
        $.get("http://qt.gtimg.cn/r=2&q=r_hk01072", function(data, status){
		  console.log(typeof(data));
        });
    });
================
This works!
$(document).ready(function(){
	$.get("http://qt.gtimg.cn/r=2&q=r_hk01072", function(data, status){console.log(data);})
});
================

linear regression

const m = tf.variable(tf.scalar(Math.random())) // slope
const b = tf.variable(tf.scalar(Math.random())) // y intercept

const predict = x =>
  tf.tidy(() => {
    const xVector = tf.tensor1d(x);    // Create a vector of x values
    const yPred = xVector.mul(this.m).add(this.b); // y = mx + b
    return yPred
  })

const loss = (yPred, y) => yPred.sub(y).square().mean(); // loss function: mean squared error

const learningRate = 0.5;
const optimizer = tf.train.sgd(learningRate);  // optimiser: stochastic gradient descent

// train function: running ~60 times per second
const train = () => { 
  tf.tidy(() => {
    if (x_vals.length > 0) {
      const y = tf.tensor1d(y_vals)
      optimiser.minimize(() => loss(predict(x_vals), y)); // auto adjusts tf.variable coefficents
    }
  })
}

// getting values from tensors is async
predict([-1, 1]).data().then(yVals => {
  two.makeLine(  // plot the Two.js line on the canvas
    // x1, y1
    -1 * width, height * yVals[0]),
    // x2, y2
    1 * width, height * yVals[1]
})

linear regression sample

<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs@0.12.0"></script>

<script>
const m = tf.variable(tf.scalar(Math.random())) // slope
const b = tf.variable(tf.scalar(Math.random())) // y intercept

function predict(x) { 
	return tf.tidy(function() { return m.mul(x).add(b);}); 
} 

function loss(prediction, actualValues) { return prediction.sub(actualValues).square().mean();} 

const learningRate = 0.01; 
const optimizer = tf.train.sgd(learningRate); 

function train() { 
	optimizer.minimize(function() { 
		const predsYs = predict(tf.tensor1d(trainX)); 
		return loss(predsYs, tf.tensor1d(trainY)) 
	}); 
} 

theStepLoss = 10
while (if theStepLoss > 0.1) {
	theStepLoss = train();
}
</script>

data.gtimg

hkHSI
http://data.gtimg.cn/flashdata/hk/minute/hkHSI.js?maxage=10
0930-1600 minute data

weekly data
http://web.ifzq.gtimg.cn/appstock/app/hkfqkline/get?_var=kline_weekqfq¶m=hk00060,week,,,320,qfq

This is monthly data!!
http://web.ifzq.gtimg.cn/appstock/app/hkfqkline/get?_var=kline_monthqfq¶m=hkHSI,month,,,320,qfq

This is Five day data!!
http://web.ifzq.gtimg.cn/appstock/app/day/query?_var=fdays_data_hkHSI&code=hkHSI

kline_day 320
http://web.ifzq.gtimg.cn/appstock/app/hkfqkline/get?_var=kline_dayqfq¶m=hkHSI,day,,,320,qfq
kline_day 30
http://web.ifzq.gtimg.cn/appstock/app/hkfqkline/get?_var=kline_dayqfq¶m=hkHSI,day,,,30,qfq

RGraph.AJAX.getJSON(url, callback)

=====================
theurl = 'http://web.ifzq.gtimg.cn/appstock/app/hkfqkline/get?_var=kline_dayqfq¶m=hkHSI,day,,,320,qfq'

RGraph.AJAX.getJSON(theurl, function (json){newjson = json});
var keys = Object.keys(newjson); // read the structure
newjson.data.hkHSI.day[320][0]  // date
newjson.data.hkHSI.day[320][1]  // open

["2018-08-22", "27836.678", "27844.932", "27919.299", "27579.801", "77596553908", {…}, "0", "7759655.391"]
date,			o		c			h		l			v					v
}