These are chat archives for harthur/brain

4th
Sep 2016
Ernani Rordrigues de SThiago
@ernanirst
Sep 04 2016 21:02

Hey guys I ran the XOR example provided and it worked fine. Then, I replaced the inputs with 360 arrays of 305 integers (I'm reading from a file of course). Each of these arrays has a label, a number from 0 to 9, which I'm using as the output.

After training this dataset (what is happening very fast and making me suspect that is not working properly), I get the result
{"error": null, "iterations": 1 }

Then I try to run the first of 360 arrays to see if I get any of the label numbers (anything from 0 to 9). Unfortunately, the result is an empty object {} which is giving me no clue about where I made a mistake.

This message was deleted
var net = new brain.NeuralNetwork();
var train = net.train(dataset);
console.log("training result: " + JSON.stringify(train));
var output = net.run(lineToArray(array[2])); 
console.log("results: " + JSON.stringify(output));
This is basically the code I'm using, if you consider that dataset is a correct training array. I can also share how I'm generating dataset.