如何使用动态数据集批量训练Brain.js神经网络？

Question

I am having hard time figuring out how to train the brain.js neural network with dynamic dataset. The GitHub documentation says the following: Each training pattern should have an input and an output, both of which can be either an array of numbers from 0 to 1 or a hash of numbers from 0 to 1

net.train([{input: [0, 0], output: [0]},
           {input: [0, 1], output: [1]},
           {input: [1, 0], output: [1]},
           {input: [1, 1], output: [0]}]);

const output = net.run([1, 0]);  // [0.987]

The problem is that I don't know beforehand how many elements are in the input array of my training data so I don't know how many {input: [0, 0], output: [0]} elements I need to pass to net.train().

For example: How do I train the neural network if I have the following arrays without hardcoding the number of {input: [0, 0], output: [0]} elements.

var input1_array = [.1, .2, .3, .4, .5]
var input2_array = [.6, .7, .8, .9, .95]
var output1_array = [.2, .6, .8, .85, .95]

// the following doesn't work 
net.train([input:[input1_array, input2_array], output:[output1_array]]);

Answer 1

I hope question is still valid.

First, I would start with blank array (ie/ input1_array=[]; ), but you can't do it as in train you need to have some value. For that reason you can put some initial data like var input1_array = [0]; if your network will have two inputs then var input1_array = [0,0];

If you will use two or more previous values as inputs you will need

var trainingdata= [{input: [0, 0], output: [0]},
           {input: [0, 1], output: [1]},
] //(sorry for typos, but idea to have two entries)

Now you have some initial zeros to not have NULL going to you NN. I was using prompt during testing and input did the following trainingdata.push (input:[prompt("Enter next in value")],output:[prompt("Enter next out value")])

Depending on what you do with the network, you may need retrain it each time a new value comes (like time series prediction with LSTM) or just collect data (NN type).

In your example replace prompt with whatever source of data like something from page ( getElementById ) or some server data JSON parsed value.

I found one trick. Make function to push to array is length

You can do something like:

if (trainingData.length < maxtrainingdata) {
  trainingData.push({input: [var1], output: [var2]});
}
else {
  trainingData.shift();
  trainingData.push({input: [var1], output: [
}