人工神经网路分析3.pptVIP

第3章 感知器 主要内容： 感知器与人工神经网络的早期发展； 线性可分问题与线性不可分问题； 感知器的训练算法。 重点： 感知器的结构、表达能力、学习算法 难点： 感知器的表达能力 第3章 感知器 3.1 感知器与人工神经网络的早期发展 3.2 感知器的学习算法 3.2.1 离散单输出感知器训练算法 3.2.2 离散多输出感知器训练算法 3.2.3 连续多输出感知器训练算法 3.3 线性不可分问题 3.3.1 异或(Exclusive –OR)问题 3.3.2 线性不可分问题的克服 3.1 感知器与ANN的早期发展 3.1 感知器与ANN的早期发展 1962年，Rosenblatt宣布：人工神经网络可以学会它能表示的任何东西 3.2 感知器的学习算法 感知器的学习是有导师学习 基本思想： 逐步地将样本集中的样本输入到网络中，根据输出结果和理想输出之间的差别来调整网络中的权值矩阵。 Training a Single-Layered Perceptron Learning Rules Linear Threshold Units (LTUs) : Perceptron Learning Rule Linearly Graded Units (LGUs) : Widrow-Hoff learning Rule 3.2.1离散单输出感知器训练算法 二值网络：自变量及其函数的值、向量分量的值只取0和1函数、向量。 权向量：W=(w1，w2，…，wn) 输入向量：X=(x1，x2，…，xn) 训练样本集： {(X，Y)|Y为输入向量X对应的输出} Perceptron Example Augmented input vector Augmented input vector Linearly Separable vs. Linearly Non-Separable Goal Given training sets T1?C1 and T2 ? C2 with elements in form of x=(x1, x2 , ... , xm-1 , xm) T , where x1, x2 , ... , xm-1 ?R and xm= ?1. Assume T1 and T2 are linearly separable. Find w=(w1, w2 , ... , wm) T such that Observation Perceptron Learning Rule Perceptron Learning Rule Summary ? Perceptron Learning Rule Summary ? Perceptron Learning Rule Perceptron Convergence Theorem The Learning Scenario The Learning Scenario The Learning Scenario The Learning Scenario The Learning Scenario The Learning Scenario The Learning Scenario Weight Space Weight Space Weight Space Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space The Learning Scenario in Weight Space Adaline (Adaptive Linear Element) LMS (Least Mean