Processing 特训（Neural Networks） - 10

明确概念

Strategies of learning: Supervised learning(有测试集结果), Unsupervised learning(无测试集结果), Reinforcement learning(有环境感知)

Perception algorithm: input by its weight, sum all of the weighted inputs, Compute the perception output

lerp: lerp(start,stop,amt), 一个变量在 start 和 stop 之间按照 amt 取值指定一个数，the lerp function is convenient for creating motion along a straight path and for drawing dotted lines.

明确功能

1.通过感知算法判断一个点落在线的左边还是右边

-Provide perception inputs and known answer

-Ask perception to guess the answer

-Compute the error

=desired - guess

=input * (desired - guess)

=weight + input * (desired - guess)

=weight + input * (desired - guess) * learning rate

-Adjust all the weights according to the error

-Repeat N times to get appropriate input weights

功能设计

类：Perception

属性：weights

方法：guess, feedforward

伪代码

setup(){

}

draw(){

}

class Perception{

    定义属性

    Perception(){}

    定义方法

}

正式代码

见：

Processing/Neural_networks at master · JoeyQiang/Processing · GitHub

代码优化

1.点集出现“虚拟线”的左右两侧

2.Neural networks diagram