Results of NeuralNetwork is unexpected

[dougmolineux]

What is wrong?

Thank you for this wonderful project. I am attempting to create a program that can tell if a person is a child or not with provided height and weight. For some reason, it is telling me the opposite results of what I would expect.

Where does it happen?

In the a NeuralNetwork when trying to the program in node.js on my mac.

How do we replicate the issue?

const brain = require("brain.js");
const net = new brain.NeuralNetwork();

net.train([
  { input: { height: 6, weight: 180 }, output: { notAChild: 1 } },
  { input: { height: 6, weight: 190 }, output: { notAChild: 1 } },
  { input: { height: 6, weight: 193 }, output: { notAChild: 1 } },
  { input: { height: 5, weight: 150 }, output: { notAChild: 1 } },
  { input: { height: 5, weight: 160 }, output: { notAChild: 1 } },
  { input: { height: 5, weight: 164 }, output: { notAChild: 1 } },
  { input: { height: 3, weight: 60 }, output: { child: 1 } },
  { input: { height: 3, weight: 68 }, output: { child: 1 } },
  { input: { height: 3, weight: 69 }, output: { child: 1 } },
])

const output = net.run({ height: 3, weight: 65 }); 
// { notAChild: 0.6541248559951782, child: 0.34587541222572327 }

console.log(output);

How important is this (1-5)?

1 - I maybe doing something wrong here.

Expected behavior (i.e. solution)

I would have expected that when given { height: 3, weight: 65 } it would have given a high likelihood that it IS a child. But it's actually saying its more likely to be a not a child. What am I doing wrong here?

Thanks again!

[nmish1995]

You have to normalize your input data. Watch this tutorial -> https://scrimba.com/p/pVZJQfg/cpyevhz

[dougmolineux]

Hmm that guy is dividing by the lowest value in the data.

[dougmolineux]

Not really sure how it applies to my use case, thanks any way. I'll close this now.