机器学习基石下 (Machine Learning Foundations)—Mathematical Foundations
Hsuan-Tien Lin, 林轩田,副教授 (Associate Professor),资讯工程学系 (Computer Science and Information Engineering)
Logistic Regression Problem
- logistic hypothesis: h(x)=θ(wTx) h ( x ) = θ ( w T x )
- logistic function θ(s)=11+e−s θ ( s ) = 1 1 + e − s
- sign(h(x)−0.5)=sign(wTx) s i g n ( h ( x ) − 0.5 ) = s i g n ( w T x )
Logistic Regression Error
Likelihood of Logistic Hypothesis
极大似然
cross-entropy error
在极大似然估计下,logistic方程的误差函数
Gradient of Logistic Regression Error
Minimizing Ein(w) E i n ( w )
PLA Revisited: Iterative Optimization
Gradient Descent
Linear Approximation
Simple Heuristic for Changing η η
Logistic Regression Algorithm