33 lines
840 B
JavaScript
33 lines
840 B
JavaScript
|
class NeuralNetwork {
|
||
|
|
constructor() {
|
||
|
|
this.nn = tf.sequential();
|
||
|
|
this.hidden1 = tf.layers.dense({
|
||
|
|
units: 6,
|
||
|
|
inputShape: [9]
|
||
|
|
});
|
||
|
|
this.hidden2 = tf.layers.dense({
|
||
|
|
units: 6
|
||
|
|
});
|
||
|
|
this.output = tf.layers.dense({
|
||
|
|
units: 2
|
||
|
|
});
|
||
|
|
this.nn.add(this.hidden1)
|
||
|
|
this.nn.add(this.hidden2);
|
||
|
|
this.nn.add(this.output);
|
||
|
|
this.nn.compile({
|
||
|
|
optimizer: tf.train.adam(1),
|
||
|
|
loss: 'meanSquaredError'
|
||
|
|
});
|
||
|
|
|
||
|
|
this.predictions = [1,0];
|
||
|
|
}
|
||
|
|
|
||
|
|
drive(inputs) {
|
||
|
|
return this.nn.predict(tf.tensor([inputs])).arraySync()[0];
|
||
|
|
}
|
||
|
|
prediction(data){
|
||
|
|
let temp = this.nn.predict(tf.tensor([data]));
|
||
|
|
this.predictions = temp.arraySync()[0];
|
||
|
|
tf.dispose(temp);
|
||
|
|
}
|
||
|
|
}
|