Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- \begin{enumerate}
- \item
- \text{Let} {A} = \left(
- \begin{array}{cccc}
- a_{11} & a_{12} & ... & a_{1m} \\
- a_{21} & a_{22} & ... & a_{2m} \\
- a_{31}& a_{32} & ... & a_{3m} \\
- \vdots & \vdots & \ddots & \vdots \\
- a_{n1} & a_{n2} & \cdots & a_{nm}\\
- \end{array}
- \right)
- \\
- \\
- \\
- \text{Let} {B} = \left(
- \begin{array}{cccc}
- b_{11} & b_{12} & ... & b_{1m} \\
- b_{21} & b_{22} & ... & b_{2m} \\
- b_{31}& b_{32} & ... & b_{3m} \\
- \vdots & \vdots & \ddots & \vdots \\
- b_{n1} & b_{n2} & \cdots & b_{nm}\\
- \end{array}
- \right).
- \\
- \\
- \text{Then}\\
- A^T + B^T = \left(
- \begin{array}{cccc}
- a_{11} + b_{11} & a_{21} + b_{21} & ... & a_{n1} + b_{n1} \\
- a_{12} + b_{12} & a_{22} + b_{22} & ... & a_{n2} + b_{n2} \\
- a_{31} + b_{13}& a_{23} + b_{23} & ... & a_{n3} + b_{n3} \\
- \vdots & \vdots & \ddots & \vdots \\
- a_{1m} + b_{1m} & au_{n2} & \cdots & a_{nm} + b_{nm}\\
- \end{array}
- \right)
- \\
- \\
- A + B = \left(\begin{array}{cccc}
- a_{11} + b_{11} & a_{21} + b_{21} & ... & a_{1m} + b_{1m} \\
- a_{21} + b_{21} & a_{22} + b_22 & ... & a_{2m} + b_{2m} \\
- a_{31} + b_{31}& a_{23} + b_{23} & ... & a_{3m} + b_{3m} \\
- \vdots & \vdots & \ddots & \vdots \\
- a_{1m} + b_{1m} & a_{n2} & \cdots & a_{nm} + b_{nm}\\
- \end{array}
- \right)
- \text{The Transpose if this sum is clearly the same as} A^T + B^T
- \item
- Let \textit{A} be an arbitrary matrix that is symmetric and invertible.\\
- A property of the identity matrix is\\
- I = I^T \\
- \textit{ Since any matrix times its inverse is the identity:}\\
- AA^{-1} = (AA^{-1})^T\\
- A^{-1}A = (A^{-1})^TA^T\\
- \textit{Since A is symmetric:} \\
- A^{-1}A = (A^{-1})^TA\\
- A^{-1}AA^{-1} = (A^{-1})^TAA^{-1}\\
- A^{-1}I = (A^{-1})^TI\\
- Therefore A^{-1} = (A^{-1})^T \textit{when A is symmetric}\\
- \item
- We know that (AB)^{-1}AB = I\\
- B^{-1}A^{-1}AB = B^{-1}IB = B^{-1}B=I\\
- \text{Since} I = (AB)^{-1}AB = B^{-1}A^{-1}AB \\
- (AB)^{-1}AB = B^{-1}A^{-1}AB \textit{if we divide by AB}
- \end{enumerate}
- \item
- \begin{enumerate}
- \item
- \begin{enumerate}
- \item
- $f(x_1,x_2,x_3)=1$ when $x_1=1, x_2=0, x_3=1$ \\
- This means that $\theta\cdot x_1 + \theta \cdot x_3 > 0$ and $\theta_1 + \theta_3 > 0$.\\
- Also, $f(x_1,x_2,x_3)=0$ when $x_1=1, x_2=0, x_3=0$ and $x_1=0, x_2=0, x_3=1$
- This suggests $\theta_1 < 0$ and $\theta_3 < 0$ respectively.\\
- This contradicts our original inequality since the sum of two negatives cannot be positive. Therefore it is not possible to learn such a $\bar{\theta}$\\
- \item
- We can simply add $b$ to each of the inequalities derived from part (i).\\
- (1) $\theta_1 + \theta_3 + b > 0$\\
- (2) $\theta_1 + b < 0$\\
- (3) $\theta_3 + b < 0$\\
- If we combine inequalities (2) and (3), we get $\theta_1 +\theta_3 < -2b$ and rearranging (1) gives us $\theta_1 + \theta_3 > -b$. Now we know that there is no solution since $\theta_1 + \theta_3$ cannot be less than $-2b$, but also greater than $-b$
- \end{enumerate}
- \item
- \begin{enumerate}
- \item Yes. If $\theta = [2,1]^T$ it correctly classifies the points.
- \item Yes. If $\theta = [2,1]^T$ and $b=-1$, the points are correctly classified.
- \item No. If the square is formed by $s$ just greater than $1$, it is the minimum size to include the point $[0,1]$, but it doesn't correctly classify the data, since $[2,-2]$ is not included. $s$ must be greater than $2$ for $[2,-2]$ to be correctly classified, but $[-2,-1]$ will be classified as positive, and thus misclassified.
- \item Yes. Centered at $[2,0]$ with $s=5$
- \end{enumerate}
- \end{enumerate}
- \end{enumerate}
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement