正文
引理1(凸的中点凸判定) (https://en.wikipedia.org/wiki/Convex_function)
凸函数的等价判定条件: ①与②等价.
① f ( x + y 2 ) ⩽ f ( x ) + f ( y ) 2 f(\frac{x+y}{2}) \leqslant \frac{f(x)+f(y)}{2} f(2x+y)⩽2f(x)+f(y)
② f ( θ x + ( 1 − θ ) y ) ⩽ θ f ( x ) + ( 1 − θ ) f ( y ) , θ ∈ [ 0 , 1 ] f({\theta}x+(1-{\theta})y) \leqslant {\theta}f(x)+(1-{\theta})f(y), {\theta}\in[0,1] f(θx+(1−θ)y)⩽θf(x)+(1−θ)f(y),θ∈[0,1].
①⇐②显然; ①⇒②对real-valued Lebesgue measurable function f f f成立, 特别的, 对continuous function f f f成立.
由于本题中都是连续函数, 因此在附录中给出了一个, 对于连续函数 f f f的, ①⇒②的证明.
引理2(凸的一阶微分判定)
下证: 若 f ( x ) ⩾ f ( y ) + ∇ f ( y ) ⊤ ( x − y ) f(x) \geqslant f(y) + \nabla f(y)^\top (x-y) f(x)⩾f(y)+∇f(y)⊤(x−y), 则 f ( θ x + ( 1 − θ ) y ) ⩾ θ f ( x ) + ( 1 − θ ) f ( y ) f({\theta}x+(1-{\theta})y) \geqslant {\theta}f(x) + (1-{\theta})f(y) f(θx+(1−θ)y)⩾θf(x)+(1−θ)f(y), 进而 f f f凸.
记 z = θ x + ( 1 − θ ) y z = {\theta}x+(1-{\theta})y z=θx+(1−θ)y, 则:
(Ⅰ) f ( x ) ⩾ f ( z ) + ∇ f ( z ) ⊤ ( x − z ) f(x) \geqslant f(z) + \nabla f(z)^\top (x-z) f(x)⩾f(z)+∇f(z)⊤(x−z);
(Ⅱ) f ( y ) ⩾ f ( z ) + ∇ f ( z ) ⊤ ( y − z ) f(y) \geqslant f(z) + \nabla f(z)^\top (y-z) f(y)⩾f(z)+∇f(z)⊤(y−z);
则: ( θ {\theta} θ(Ⅰ)+ ( 1 − θ ) (1-{\theta}) (1−θ)(Ⅱ)): θ f ( x ) + ( 1 − θ ) f ( y ) ⩾ f ( z ) + ∇ f ( z ) ⊤ ( θ x + ( 1 − θ ) y − z ) = f ( z ) + ∇ f ( z ) ⊤ 0 = f ( z ) = f ( θ x + ( 1 − θ ) y ) {\theta}f(x) + (1-{\theta})f(y) \geqslant f(z) + \nabla f(z)^\top ({\theta}x+(1-{\theta})y - z) = f(z) + \nabla f(z)^\top 0 = f(z) = f({\theta}x+(1-{\theta})y) θf(x)+(1−θ)f(y)⩾f(z)+∇f(z)⊤(θx+(1−θ)y−z)=f(z)+∇f(z)⊤0=f(z)=f(θx+(1−θ)y)
引理3(凸的二阶微分判定)
若 ∇ 2 f ( z ) ≽ 0 \nabla^2 f(z) \succcurlyeq 0 ∇2f(z)≽0, 则 f ( x ) = f ( y ) + ∇ f ( y ) ⊤ ( x − y ) + ( x − y ) ⊤ ∇ 2 f ( ξ ) ( x − y ) ⩾ f ( y ) + ∇ f ( y ) ⊤ ( x − y ) f(x) = f(y) + \nabla f(y)^\top (x-y) + (x-y)^\top \nabla^2 f(\xi) (x-y) \geqslant f(y) + \nabla f(y)^\top (x-y) f(x)=f(y)+∇f(y)⊤(x−y)+(x−y)⊤∇2f(ξ)(x−y)⩾f(y)+∇f(y)⊤(x−y), 进而 f f f凸.
附录
不妨 x < y , θ ∈ ( 0 , 1 ) x < y, {\theta}\in(0,1) x<y,θ∈(0,1), 构造数列 { x i } i = 0 ∞ , { y i } i = 0 ∞ , { m i } i = 0 ∞ \{x_i\}_{i=0}^{\infty},\{y_i\}_{i=0}^{\infty},\{m_i\}_{i=0}^{\infty} { x