ESL作业笔记2.1.

本文章基本是对https://stats.stackexchange.com/questions/130998/explanation-of-formula-for-median-closest-point-to-origin-of-n-samples-from-unit 的翻译。
解决问题为Element of Statistical Learning (ESL, https://web.stanford.edu/~hastie/ElemStatLearn/)课后习题2.1.

---

题目简介

Suppose each of $K$-classes has an associated target $t_k$, which is a vector of all zeros, except a one in the kth position. Show that classifying to the largest element of $\hat y$ amounts to choosing the closest target, $\min_k ||t_k-\hat y||^2$, if the elements of $\hat y$ sum to one.

解题

$$||t_k-\hat y||^2 = \sum_{i=1}^K\hat y_i^2 - \hat y_k^2 + (1-\hat y_k)^2 = \sum_{i=1}^K\hat y_i^2 + 1 - 2\hat y_k\\ \text{argmin}_k ||t_k-\hat y||^2 = \text{argmin}(-2\hat y_k) = \text{argmax}_k (\hat y_k)$$