0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?

線形判別関数で使用する線形代数の知識

Last updated at Posted at 2024-12-29

2変数線形判別分析では, 群間変動$w^{T}S_Bw$を大きく, 群内変動$w^{T}S_Ww$を小さくしたい. つまり以下の$\lambda$の最大化を目指す.

\lambda = \frac{w^{T}S_{B}w}{w^{T}S_ww}

ここで, $S_B$は半正定値行列, $S_W$は正定値行列より, $\lambda \ge 0$.

$w^{T}S_ww=1$のもと, ラグランジュの未定乗数法を用いると,

S_W^{-1}S_Bw = \lambda w

となり, $S_W^{-1}S_B$の最大固有値を求めればよいことがわかる.
ここで, $S_B$のランクは1より, $S_W^{-1}S_B$のランクも1となる.
行列のランク$\ge$非ゼロ固有値の数 より, $S_W^{-1}S_B$は最大で1つの非ゼロ固有値$(>0)$を持つ. よって, $S_W^{-1}S_B$における非ゼロ固有値に対応する固有ベクトル$\hat{w}$を求めればよい.

行列のランク$\ge$非ゼロ固有値の数 は, 任意の正方行列Aが正則行列Pを用いて三角化できることから確認できる($\Sigma$:上三角行列).

P^{-1}AP = \Sigma

$rank(左辺) = rank(P^{-1}AP)=rank(A) =rank(\Sigma) = rank(右辺)$ .

0
0
0

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
0
0

Delete article

Deleted articles cannot be recovered.

Draft of this article would be also deleted.

Are you sure you want to delete this article?