โ€œMachine Learningโ€์„ ๊ณต๋ถ€ํ•˜๋ฉด์„œ ๊ฐœ์ธ์ ์ธ ์šฉ๋„๋กœ ์ •๋ฆฌํ•œ ํฌ์ŠคํŠธ์ž…๋‹ˆ๋‹ค. ์ง€์ ์€ ์–ธ์ œ๋‚˜ ํ™˜์˜์ž…๋‹ˆ๋‹ค :)

1 minute read

โ€œMachine Learningโ€์„ ๊ณต๋ถ€ํ•˜๋ฉด์„œ ๊ฐœ์ธ์ ์ธ ์šฉ๋„๋กœ ์ •๋ฆฌํ•œ ํฌ์ŠคํŠธ์ž…๋‹ˆ๋‹ค. ์ง€์ ์€ ์–ธ์ œ๋‚˜ ํ™˜์˜์ž…๋‹ˆ๋‹ค :)


Definition. basis function; ๊ธฐ์ € ํ•จ์ˆ˜

$y(\mathbf{x}, \mathbf{w}) = \mathbf{w} \cdot \mathbf{x}$์ธ Linear Regression์‹์— input $\mathbf{x}$๋ฅผ transform ํ•˜๋Š” ํ•จ์ˆ˜๋ฅผ ๋งํ•จ.

\[y(\mathbf{x}, \mathbf{w}) = \mathbf{w} \cdot \phi(\mathbf{x})\]

์ด๋Ÿฐ basis function ์—ฌ๋Ÿฌ ๊ฐœ๋ฅผ ์‚ฌ์šฉํ•ด Linear Regression1 ํ•  ์ˆ˜๋„ ์žˆ๋‹ค.

\[y(\mathbf{x}, \mathbf{w}) = \sum_{j=1}^{M} \mathbf{w}_j \cdot \phi_j(\mathbf{x})\]

Example. basis function

  • polynomial basis function
\[\phi_j(x) = (x)^j\]
  • sigmoid basis function
\[\begin{aligned} \phi_j(x) = \sigma \left( \frac{x - u_j}{s}\right) \\ \sigma(a) = 1 / (1 + \exp (-a)) \end{aligned}\]
  • gaussian basis function
\[\phi_j(x) = \exp \left( - \frac{(x - u_j)^2}{2s^2}\right)\]

references


  1. $x$์— ๋Œ€ํ•ด์„œ๋Š” ๋น„์„ ํ˜• ํ•จ์ˆ˜์ด์ง€๋งŒ, $w$์— ๋Œ€ํ•ด์„œ ์„ ํ˜•์ด๋ฏ€๋กœ Linear Regression์ด๋ผ๊ณ  ๋ถ€๋ฅธ๋‹ค.ย