18. Linear Programming
Ax = b
x 0
Objective function:
minimize
Constraint:
z = cT
x
X
y
(x, y) = Pr(x)
X
x
(x, y) = P✓(y)
Objective function:
Constraint:
8x, y (x, y) 0
EMD(Pr, P✓) = inf
2⇧(Pr,P✓)
hD, iF
21. Dual Form
Ax = b
x 0
z = cT
x ˜z = bT
y
z = cT
x yT
Ax = yT
b = ˜z
AT
y c
z = ˜zStrong Duality:
Weak Duality: z ˜z
Primal Problem: Dual Problem:
minimize: maximize:
constraint: constraint:
is a lower bound of z = cT
x yT
Ax = yT
bx yT
Ax = yT
b = ˜z
37. About the Speaker
Mark Chang
• Email: ckmarkoh at gmail dot com
• Blog: https://ckmarkoh.github.io/
• Github: https://github.com/ckmarkoh
• Slideshare: http://www.slideshare.net/ckmarkohchang
• Youtube: https://www.youtube.com/channel/UCckNPGDL21aznRhl3EijRQw
37
HTC Research & Healthcare
Deep Learning Algorithms
Research Engineer