最近在跟台大的这个课程,觉得不错,想把学习笔记发出来跟大家分享下,有错误希望大家指正. 一机器学习是什么? 感觉和 Tom M. Mitchell的定义几乎一致, A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by
for batch&supervised binary classfication,g≈f <=> Eout(g)≥0 achieved through Eout(g)≈Ein(g) and Ein(g)≈0 其中Ein是某一个备选函数h在数据D上犯错误的比例,在整个数据集上犯错误的比例为Eout 1.Perceptron Hypothesis Set 假设训数据集市线性可分的,感知机学习是目标就是求得一个能够将训练集正实例点和负实例点完全正确分开的分离超平面, 对于一组数据X={x1