javascript - NOT 函数的 bool 感知器

标签 javascript machine-learning neural-network boolean-logic perceptron

我正在尝试为基本 bool 表达式实现一个简单的感知器。但我无法正确训练 NOT 感知器。

我成功地训练了AND和OR感知器以返回给定输入集的正确值。但是当我尝试训练NOT时。

我就是这样做的:

AND 和 OR 感知器有两个输入、两个权重和一个偏差(固定 1 为偏置输入)。

所有感知器的所有权重均以0开头。 然后，我生成随机值(0 到 1 之间)来训练感知器，并使其保持循环，直到获得 10 次正确的猜测。

他们的学习率为0.1

这是训练过程:

猜测值:
对于每个输入，我将输入乘以权重，并对所有值(包括偏差)求和。

sum = (weight1 * input1) + (weight2 * input2) + (biasWeight * biasInput)--Bias input is fixed to 1
return = if (sum > 0) then 1 else 0

训练感知器:
我从感知器得到猜测

val = and.guess(1,0) --This will return 0 or 1
error = answer - val

对于每个输入，我都会执行此计算

weight = weight + (input * error * rate)

然后我对偏差做同样的事情

biasWeight = biasWeight + (input * error * rate)--Bias input is fixed to 1

通过这个过程，我可以成功训练AND和OR感知器。

AND/OR 和 NOT 感知器之间的唯一区别是输入的数量(仅1 表示不)

但是 NOT 感知器只是根据学习率中的数字不断增加权重。

有时，根据NOT感知器的训练顺序，当感知器达到0.5时，它会获得正确的值。

当我回家发布代码时，有代码(html，javascript)。我确实发现了这个bug。应该返回权重*输入的CALC函数正在返回权重+输入，并且它实际上适用于AND和或培训。

<!DOCTYPE html>

<html lang="en" xmlns="http://www.w3.org/1999/xhtml">
<head>
    <meta charset="utf-8" />
    <title></title>
    <script src="jquery-3.2.1.js"></script>

    <script type="text/javascript">
        function Show(text) {
            if (!text) {
                text = '';
            }

            document.writeln(text + '<br />');
        }

        //return random value from 0 ~ 1
        function getRandom() {
            return Math.floor(Math.random() * 2);
        };

        function PerceptronData(input, weight) {
            this.input = input;
            this.weight = weight;
        }
        PerceptronData.prototype.calc = function () {
            var result = this.input + this.weight;
            return result;
        };
        PerceptronData.prototype.adjust = function (error, rate) {
            this.weight += (this.input * error * rate);
        };
        PerceptronData.prototype.print = function () {
            return '(' + this.input + ', ' + this.weight + ')';
        }

        function Perceptron(n) {
            this.data = [];//Data array [input, weight]
            this.bias = new PerceptronData(1, 0);
            this.rate = 0.1;//learning rate

            //initial data
            for (var index = 0; index < n; index++) {
                this.data.push(new PerceptronData(0, 0));
            }
        }
        //called from "guess" function in the final perceptron
        Perceptron.prototype.process = function (inputs) {
            var data = this.data;

            if (inputs.length != data.length) {
                throw "The number os inputs [" + inputs.length + "] doesn't match with the start value [" + data.length + "] of the Perceptron.";
            }

            var dataSum = 0;
            for (var index = 0; index < data.length; index++) {
                data[index].input = parseInt(inputs[index]);
                dataSum += data[index].calc();
            }

            dataSum += this.bias.calc();

            return dataSum;
        };
        //twick the weight for every data
        Perceptron.prototype.adjust = function (value, answer) {
            var data = this.data;
            var error = answer - value;

            for (var index = 0; index < data.length; index++) {
                data[index].adjust(error, this.rate);
            }

            this.bias.adjust(error, this.rate);
        };
        Perceptron.prototype.print = function () {
            var data = this.data;
            var result = '';
            for (var index = 0; index < data.length; index++) {
                result += 'data[' + index + ']' + data[index].print() + ' > ';
            }

            return result + 'bias' + this.bias.print();
        };

        function NotPerceptron() {
            Perceptron.call(this, 1);
        }
        NotPerceptron.prototype = Object.create(Perceptron.prototype);
        NotPerceptron.prototype.guess = function (value) {
            var data = this.process([value]);

            //activation function
            return ((data > 0) ? 1 : 0);
        };
        NotPerceptron.prototype.train = function (value, answer) {
            var result = this.guess([value]);
            this.adjust(result, answer);
        };

        function AndPerceptron() {
            Perceptron.call(this, 2);
        }
        AndPerceptron.prototype = Object.create(Perceptron.prototype);
        AndPerceptron.prototype.guess = function (valueA, valueB) {
            var data = this.process([valueA, valueB]);

            //activation function
            return ((data > 0) ? 1 : 0);
        };
        AndPerceptron.prototype.train = function (valueA, valueB, answer) {
            var result = this.guess(valueA, valueB);

            this.adjust(result, answer);
        };

        function OrPerceptron() {
            Perceptron.call(this, 2);
        }
        OrPerceptron.prototype = Object.create(Perceptron.prototype);
        OrPerceptron.prototype.guess = function (valueA, valueB) {
            var data = this.process([valueA, valueB]);

            //activation function
            return ((data > 0) ? 1 : 0);
        };
        OrPerceptron.prototype.train = function (valueA, valueB, answer) {
            var result = this.guess(valueA, valueB);

            this.adjust(result, answer);
        };
    </script>
</head>
<body>
    <script type="text/javascript">
        Show('Training AND...');
        Show();
        var and = new AndPerceptron();

        var count = 0;
        var total = 0;
        var max = 100;

        while (count < 10 && total < max) {
            total++;
            var a = getRandom();
            var b = getRandom();
            var answer = ((a === 1 && b === 1) ? 1 : 0);

            and.train(a, b, answer);

            a = getRandom();
            b = getRandom();
            answer = ((a === 1 && b === 1) ? 1 : 0);

            var guess = and.guess(a, b);

            if (guess === answer) {
                count++;
            } else {
                count = 0;
            }

            Show(' > AND(' + a + ', ' + b + ') = ' + guess + ' > [' + and.print() + ']');

            if (count == 10) {
                //final test
                if (and.guess(0, 0) == 1) {
                    count = 0;
                }

                if (and.guess(0, 1) == 1) {
                    count = 0;
                }

                if (and.guess(1, 0) == 1) {
                    count = 0;
                }

                if (and.guess(1, 1) == 0) {
                    count = 0;
                }
            }
        }
        Show();

        if (total >= max) {
            Show('AND training failed...');
        } else {
            Show('AND trained with [' + total + '] interactions. [' + and.print() + ']');
        }

        Show();
        Show('AND(0, 0) = ' + and.guess(0, 0));
        Show('AND(0, 1) = ' + and.guess(0, 1));
        Show('AND(1, 0) = ' + and.guess(1, 0));
        Show('AND(1, 1) = ' + and.guess(1, 1));

        Show();
        Show('Training OR...');
        Show();
        var or = new OrPerceptron();

        count = 0;
        total = 0;
        max = 100;

        while (count < 10 && total < max) {
            total++;
            var a = getRandom();
            var b = getRandom();
            var answer = ((a === 1 || b === 1) ? 1 : 0);

            or.train(a, b, answer);

            a = getRandom();
            b = getRandom();
            answer = ((a === 1 || b === 1) ? 1 : 0);

            var guess = or.guess(a, b);

            if (guess === answer) {
                count++;
            } else {
                count = 0;
            }

            Show(' > OR(' + a + ', ' + b + ') = ' + guess + ' > [' + or.print() + ']');

            if (count == 10) {
                //final test
                if (or.guess(0, 0) == 1) {
                    count = 0;
                }

                if (or.guess(0, 1) == 0) {
                    count = 0;
                }

                if (or.guess(1, 0) == 0) {
                    count = 0;
                }

                if (or.guess(1, 1) == 0) {
                    count = 0;
                }
            }
        }
        Show();

        if (total >= max) {
            Show('OR training failed...');
        } else {
            Show('OR trained with [' + total + '] interactions. [' + or.print() + ']');
        }

        Show();
        Show('OR(0, 0) = ' + or.guess(0, 0));
        Show('OR(0, 1) = ' + or.guess(0, 1));
        Show('OR(1, 0) = ' + or.guess(1, 0));
        Show('OR(1, 1) = ' + or.guess(1, 1));

        Show();
        Show('Training NOT...');
        Show();
        var not = new NotPerceptron();
        not.rate = 0.1;

        count = 0;
        total = 0;
        max = 100;

        while (count < 10 && total < max) {
            total++;
            var test = getRandom();
            var answer = ((test === 1) ? 0 : 1);

            not.train(test, answer);

            test = getRandom();
            answer = ((test === 1) ? 0 : 1);

            var guess = not.guess(test);

            if (guess === answer) {
                count++;
            } else {
                count = 0;
            }

            Show(' > NOT(' + test + ') = ' + guess + ' > [' + not.print() + ']');

            if (count == 10) {
                //final test
                if (not.guess(0) == 0) {
                    count = 0;
                }

                if (not.guess(1) == 1) {
                    count = 0;
                }
            }
        }
        Show();

        if (total >= max) {
            Show('NOT training failed...');
        } else {
            Show('NOT trained with [' + total + '] interactions. [' + not.print() + ']');
        }

        Show();
        Show('NOT(1) = ' + not.guess(1));
        Show('NOT(0) = ' + not.guess(0));
    </script>
</body>
</html>

输出:

Training AND...

> AND(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, 0.1)]
> AND(1, 1) = 1 > [data[0](1, 0.1) > data[1](1, 0) > bias(1, 0)]
> AND(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0) > bias(1, 0)]
> AND(1, 1) = 1 > [data[0](1, 0.1) > data[1](1, 0) > bias(1, 0)]
> AND(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0) > bias(1, 0)]
> AND(0, 1) = 0 > [data[0](0, 0.1) > data[1](1, 0) > bias(1, 0)]
> AND(0, 1) = 0 > [data[0](0, 0) > data[1](1, 0) > bias(1, -0.1)]
> AND(0, 1) = 1 > [data[0](0, 0.1) > data[1](1, 0.1) > bias(1, 0)]
> AND(0, 1) = 0 > [data[0](0, 0.1) > data[1](1, 0) > bias(1, -0.1)]
> AND(1, 1) = 0 > [data[0](1, 0.1) > data[1](1, 0) > bias(1, -0.1)]
> AND(1, 1) = 0 > [data[0](1, 0.1) > data[1](1, 0) > bias(1, -0.1)]
> AND(1, 0) = 0 > [data[0](1, 0.1) > data[1](0, 0) > bias(1, -0.1)]
> AND(1, 1) = 1 > [data[0](1, 0.2) > data[1](1, 0.1) > bias(1, 0)]
> AND(0, 0) = 0 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]
> AND(1, 0) = 0 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]
> AND(1, 0) = 0 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]
> AND(1, 0) = 0 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]
> AND(1, 0) = 0 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]
> AND(0, 0) = 0 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]
> AND(1, 0) = 0 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]
> AND(0, 0) = 0 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, -0.1)]

AND trained with [21] interactions. [data[0](1, 0.1) > data[1](1, 0.1) > bias(1, -0.1)]

AND(0, 0) = 0
AND(0, 1) = 0
AND(1, 0) = 0
AND(1, 1) = 1

Training OR...

> OR(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, 0.1)]
> OR(0, 1) = 1 > [data[0](0, 0.1) > data[1](1, 0.1) > bias(1, 0.1)]
> OR(0, 1) = 1 > [data[0](0, 0.1) > data[1](1, 0.1) > bias(1, 0.1)]
> OR(0, 0) = 1 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, 0.1)]
> OR(0, 0) = 1 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, 0.1)]
> OR(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, 0.1)]
> OR(0, 1) = 1 > [data[0](0, 0.1) > data[1](1, 0.1) > bias(1, 0.1)]
> OR(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, 0)]
> OR(0, 0) = 0 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, 0)]
> OR(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, 0)]
> OR(0, 0) = 0 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, 0)]
> OR(0, 0) = 0 > [data[0](0, 0.1) > data[1](0, 0.1) > bias(1, 0)]
> OR(1, 1) = 1 > [data[0](1, 0.1) > data[1](1, 0.1) > bias(1, 0)]
> OR(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, 0)]
> OR(1, 0) = 1 > [data[0](1, 0.1) > data[1](0, 0.1) > bias(1, 0)]

OR trained with [15] interactions. [data[0](1, 0.1) > data[1](1, 0.1) > bias(1, 0)]

OR(0, 0) = 0
OR(0, 1) = 1
OR(1, 0) = 1
OR(1, 1) = 1

Training NOT...

> NOT(0) = 0 > [data[0](0, 0) > bias(1, 0)]
> NOT(1) = 1 > [data[0](1, 0) > bias(1, 0.1)]
> NOT(0) = 1 > [data[0](0, 0) > bias(1, 0.1)]
> NOT(1) = 1 > [data[0](1, 0) > bias(1, 0.1)]
> NOT(0) = 0 > [data[0](0, -0.1) > bias(1, 0)]
> NOT(1) = 1 > [data[0](1, -0.2) > bias(1, -0.1)]
> NOT(1) = 1 > [data[0](1, -0.2) > bias(1, -0.1)]
> NOT(0) = 1 > [data[0](0, -0.2) > bias(1, -0.1)]
> NOT(0) = 1 > [data[0](0, -0.30000000000000004) > bias(1, -0.2)]
> NOT(1) = 1 > [data[0](1, -0.30000000000000004) > bias(1, -0.2)]
> NOT(0) = 1 > [data[0](0, -0.30000000000000004) > bias(1, -0.2)]
> NOT(1) = 1 > [data[0](1, -0.4) > bias(1, -0.30000000000000004)]
> NOT(1) = 1 > [data[0](1, -0.5) > bias(1, -0.4)]
> NOT(1) = 1 > [data[0](1, -0.5) > bias(1, -0.4)]
> NOT(1) = 1 > [data[0](1, -0.6) > bias(1, -0.5)]
> NOT(1) = 1 > [data[0](1, -0.6) > bias(1, -0.5)]
> NOT(1) = 1 > [data[0](1, -0.7) > bias(1, -0.6)]
> NOT(1) = 1 > [data[0](1, -0.7999999999999999) > bias(1, -0.7)]
> NOT(0) = 1 > [data[0](0, -0.8999999999999999) > bias(1, -0.7999999999999999)]
> NOT(0) = 1 > [data[0](0, -0.8999999999999999) > bias(1, -0.7999999999999999)]
> NOT(0) = 1 > [data[0](0, -0.9999999999999999) > bias(1, -0.8999999999999999)]
> NOT(0) = 1 > [data[0](0, -0.9999999999999999) > bias(1, -0.8999999999999999)]
> NOT(1) = 1 > [data[0](1, -0.9999999999999999) > bias(1, -0.8999999999999999)]
> NOT(0) = 1 > [data[0](0, -0.9999999999999999) > bias(1, -0.8999999999999999)]
> NOT(0) = 1 > [data[0](0, -1.0999999999999999) > bias(1, -0.9999999999999999)]
> NOT(1) = 1 > [data[0](1, -1.2) > bias(1, -1.0999999999999999)]
> NOT(0) = 1 > [data[0](0, -1.2) > bias(1, -1.0999999999999999)]
> NOT(1) = 1 > [data[0](1, -1.2) > bias(1, -1.0999999999999999)]
> NOT(0) = 1 > [data[0](0, -1.2) > bias(1, -1.0999999999999999)]
> NOT(0) = 1 > [data[0](0, -1.2) > bias(1, -1.0999999999999999)]
> NOT(1) = 1 > [data[0](1, -1.2) > bias(1, -1.0999999999999999)]
> NOT(1) = 1 > [data[0](1, -1.3) > bias(1, -1.2)]
> NOT(0) = 1 > [data[0](0, -1.4000000000000001) > bias(1, -1.3)]
> NOT(0) = 1 > [data[0](0, -1.5000000000000002) > bias(1, -1.4000000000000001)]
> NOT(1) = 1 > [data[0](1, -1.6000000000000003) > bias(1, -1.5000000000000002)]
> NOT(1) = 1 > [data[0](1, -1.6000000000000003) > bias(1, -1.5000000000000002)]
> NOT(0) = 1 > [data[0](0, -1.6000000000000003) > bias(1, -1.5000000000000002)]
> NOT(0) = 1 > [data[0](0, -1.7000000000000004) > bias(1, -1.6000000000000003)]
> NOT(0) = 1 > [data[0](0, -1.8000000000000005) > bias(1, -1.7000000000000004)]
> NOT(1) = 1 > [data[0](1, -1.9000000000000006) > bias(1, -1.8000000000000005)]
> NOT(1) = 1 > [data[0](1, -1.9000000000000006) > bias(1, -1.8000000000000005)]
> NOT(1) = 1 > [data[0](1, -1.9000000000000006) > bias(1, -1.8000000000000005)]
> NOT(1) = 1 > [data[0](1, -1.9000000000000006) > bias(1, -1.8000000000000005)]
> NOT(0) = 1 > [data[0](0, -2.0000000000000004) > bias(1, -1.9000000000000006)]
> NOT(1) = 1 > [data[0](1, -2.1000000000000005) > bias(1, -2.0000000000000004)]
> NOT(1) = 1 > [data[0](1, -2.2000000000000006) > bias(1, -2.1000000000000005)]
> NOT(1) = 1 > [data[0](1, -2.3000000000000007) > bias(1, -2.2000000000000006)]
> NOT(0) = 1 > [data[0](0, -2.3000000000000007) > bias(1, -2.2000000000000006)]
> NOT(0) = 1 > [data[0](0, -2.400000000000001) > bias(1, -2.3000000000000007)]
> NOT(0) = 1 > [data[0](0, -2.500000000000001) > bias(1, -2.400000000000001)]
> NOT(1) = 1 > [data[0](1, -2.600000000000001) > bias(1, -2.500000000000001)]
> NOT(0) = 1 > [data[0](0, -2.700000000000001) > bias(1, -2.600000000000001)]
> NOT(1) = 1 > [data[0](1, -2.800000000000001) > bias(1, -2.700000000000001)]
> NOT(0) = 1 > [data[0](0, -2.9000000000000012) > bias(1, -2.800000000000001)]
> NOT(1) = 1 > [data[0](1, -3.0000000000000013) > bias(1, -2.9000000000000012)]
> NOT(1) = 1 > [data[0](1, -3.0000000000000013) > bias(1, -2.9000000000000012)]
> NOT(1) = 1 > [data[0](1, -3.0000000000000013) > bias(1, -2.9000000000000012)]
> NOT(0) = 1 > [data[0](0, -3.1000000000000014) > bias(1, -3.0000000000000013)]
> NOT(0) = 1 > [data[0](0, -3.1000000000000014) > bias(1, -3.0000000000000013)]
> NOT(1) = 1 > [data[0](1, -3.2000000000000015) > bias(1, -3.1000000000000014)]
> NOT(0) = 1 > [data[0](0, -3.3000000000000016) > bias(1, -3.2000000000000015)]
> NOT(1) = 1 > [data[0](1, -3.4000000000000017) > bias(1, -3.3000000000000016)]
> NOT(0) = 1 > [data[0](0, -3.5000000000000018) > bias(1, -3.4000000000000017)]
> NOT(0) = 1 > [data[0](0, -3.600000000000002) > bias(1, -3.5000000000000018)]
> NOT(1) = 1 > [data[0](1, -3.700000000000002) > bias(1, -3.600000000000002)]
> NOT(1) = 1 > [data[0](1, -3.700000000000002) > bias(1, -3.600000000000002)]
> NOT(1) = 1 > [data[0](1, -3.800000000000002) > bias(1, -3.700000000000002)]
> NOT(0) = 1 > [data[0](0, -3.800000000000002) > bias(1, -3.700000000000002)]
> NOT(1) = 1 > [data[0](1, -3.900000000000002) > bias(1, -3.800000000000002)]
> NOT(1) = 1 > [data[0](1, -4.000000000000002) > bias(1, -3.900000000000002)]
> NOT(1) = 1 > [data[0](1, -4.000000000000002) > bias(1, -3.900000000000002)]
> NOT(0) = 1 > [data[0](0, -4.000000000000002) > bias(1, -3.900000000000002)]
> NOT(0) = 1 > [data[0](0, -4.000000000000002) > bias(1, -3.900000000000002)]
> NOT(1) = 1 > [data[0](1, -4.100000000000001) > bias(1, -4.000000000000002)]
> NOT(1) = 1 > [data[0](1, -4.100000000000001) > bias(1, -4.000000000000002)]
> NOT(1) = 1 > [data[0](1, -4.200000000000001) > bias(1, -4.100000000000001)]
> NOT(0) = 1 > [data[0](0, -4.300000000000001) > bias(1, -4.200000000000001)]
> NOT(1) = 1 > [data[0](1, -4.300000000000001) > bias(1, -4.200000000000001)]
> NOT(1) = 1 > [data[0](1, -4.4) > bias(1, -4.300000000000001)]
> NOT(0) = 1 > [data[0](0, -4.5) > bias(1, -4.4)]
> NOT(0) = 1 > [data[0](0, -4.5) > bias(1, -4.4)]
> NOT(0) = 1 > [data[0](0, -4.5) > bias(1, -4.4)]
> NOT(0) = 1 > [data[0](0, -4.6) > bias(1, -4.5)]
> NOT(1) = 1 > [data[0](1, -4.699999999999999) > bias(1, -4.6)]
> NOT(0) = 1 > [data[0](0, -4.799999999999999) > bias(1, -4.699999999999999)]
> NOT(1) = 1 > [data[0](1, -4.799999999999999) > bias(1, -4.699999999999999)]
> NOT(0) = 1 > [data[0](0, -4.899999999999999) > bias(1, -4.799999999999999)]
> NOT(0) = 1 > [data[0](0, -4.999999999999998) > bias(1, -4.899999999999999)]
> NOT(0) = 1 > [data[0](0, -5.099999999999998) > bias(1, -4.999999999999998)]
> NOT(0) = 1 > [data[0](0, -5.1999999999999975) > bias(1, -5.099999999999998)]
> NOT(0) = 1 > [data[0](0, -5.299999999999997) > bias(1, -5.1999999999999975)]
> NOT(0) = 1 > [data[0](0, -5.299999999999997) > bias(1, -5.1999999999999975)]
> NOT(0) = 1 > [data[0](0, -5.299999999999997) > bias(1, -5.1999999999999975)]
> NOT(1) = 1 > [data[0](1, -5.299999999999997) > bias(1, -5.1999999999999975)]
> NOT(0) = 1 > [data[0](0, -5.299999999999997) > bias(1, -5.1999999999999975)]
> NOT(0) = 1 > [data[0](0, -5.299999999999997) > bias(1, -5.1999999999999975)]
> NOT(0) = 1 > [data[0](0, -5.399999999999997) > bias(1, -5.299999999999997)]
> NOT(0) = 1 > [data[0](0, -5.4999999999999964) > bias(1, -5.399999999999997)]
> NOT(1) = 1 > [data[0](1, -5.599999999999996) > bias(1, -5.4999999999999964)]
> NOT(0) = 1 > [data[0](0, -5.699999999999996) > bias(1, -5.599999999999996)]
> NOT(1) = 1 > [data[0](1, -5.799999999999995) > bias(1, -5.699999999999996)]
> NOT(0) = 1 > [data[0](0, -5.899999999999995) > bias(1, -5.799999999999995)]
> NOT(0) = 1 > [data[0](0, -5.999999999999995) > bias(1, -5.899999999999995)]
> NOT(0) = 1 > [data[0](0, -6.099999999999994) > bias(1, -5.999999999999995)]
> NOT(1) = 1 > [data[0](1, -6.199999999999994) > bias(1, -6.099999999999994)]
> NOT(0) = 1 > [data[0](0, -6.199999999999994) > bias(1, -6.099999999999994)]
> NOT(1) = 1 > [data[0](1, -6.199999999999994) > bias(1, -6.099999999999994)]
> NOT(1) = 1 > [data[0](1, -6.199999999999994) > bias(1, -6.099999999999994)]
> NOT(0) = 1 > [data[0](0, -6.199999999999994) > bias(1, -6.099999999999994)]
> NOT(1) = 1 > [data[0](1, -6.199999999999994) > bias(1, -6.099999999999994)]
> NOT(0) = 1 > [data[0](0, -6.199999999999994) > bias(1, -6.099999999999994)]
> NOT(1) = 1 > [data[0](1, -6.299999999999994) > bias(1, -6.199999999999994)]
> NOT(0) = 1 > [data[0](0, -6.399999999999993) > bias(1, -6.299999999999994)]

最佳答案

根据@Stanislav Kralin 的建议，我再次更新了问题，以便它显示问题。这是解决方案。

问题出在 CALC 函数上，该函数应该乘以权重的输入值。但我正在添加它。

不幸的是，我太关注是否应该使用sigmoid函数或其他函数，查看学习率以及线性和非线性函数，以至于我没有看到这个错误。

与 AND 和 OR 感知器配合良好的事实确实把我引向了错误的方向。

PerceptronData.prototype.calc = function () {
    //var result = this.input + this.weight;//This was wrong... :(
    var result = this.input * this.weight;
    return result;
};

关于javascript - NOT 函数的 bool 感知器，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/45640357/

上一篇：javascript - AppComponent.html :1 ERROR Error: No provider for Http! 在注入(inject)错误(core.es5.js:1169)

下一篇：javascript - 如何将 .scss 文件作为全局导入到 app.js 父文件？

相关文章：

javascript - 如何通过子节点递归并输出为json？

javascript - 动态更改 svg 的颜色并在 Three.js 中应用为纹理

tensorflow - 如何使用功能来评估不用于训练模型的自定义 TensorFlow 指标

java - Neuroph Studio 2.92 未显示 Neuroph 项目选项

python - 神经网络只预测一类，所有用于测试的输出层激活变成0.999~

javascript - 获取具有特定类名的每个 li 值

java - 浏览 JavaScript

solr - 抓取网页后识别产品，进行价格比较

python - 为什么PCA的数量改变了？

keras - pytorch 的交叉熵损失与 keras 的 "categorical_crossentropy"不同吗？

©2024 IT工具网  联系我们