Reputation
555
Next privilege 1,000 Rep.
Create new tags
Badges
2 14
Impact
~8k people reached

11h
accepted Making sense of $ f(y) - f(x) = \int_{\tau = 0}^{1} \langle \nabla f( x+ \tau (y - x)), y - x \rangle d \tau $
13h
revised Making sense of $ f(y) - f(x) = \int_{\tau = 0}^{1} \langle \nabla f( x+ \tau (y - x)), y - x \rangle d \tau $
edited title
13h
asked Making sense of $ f(y) - f(x) = \int_{\tau = 0}^{1} \langle \nabla f( x+ \tau (y - x)), y - x \rangle d \tau $
Aug
30
accepted Prove that $ \sum_{k=1}^T t_k f(x_k) \leq B \Rightarrow \min_{ k \in \{1, \ldots, T \} } f(x_k) \leq \frac{ B }{ \sum_{k=1}^T t_k } $
Aug
30
comment Prove that $ \sum_{k=1}^T t_k f(x_k) \leq B \Rightarrow \min_{ k \in \{1, \ldots, T \} } f(x_k) \leq \frac{ B }{ \sum_{k=1}^T t_k } $
But here your have a coefficient $t_k$, which might be different from other $t_{k^\prime}$ for $k^\prime \neq k $.
Aug
30
asked Prove that $ \sum_{k=1}^T t_k f(x_k) \leq B \Rightarrow \min_{ k \in \{1, \ldots, T \} } f(x_k) \leq \frac{ B }{ \sum_{k=1}^T t_k } $
Aug
30
comment Prove $2a^Tb \leq \|a \|^2 + \|b\|_*^2$ with dual norm
Ah ... right ...
Aug
30
accepted Prove $2a^Tb \leq \|a \|^2 + \|b\|_*^2$ with dual norm
Aug
30
asked Prove $2a^Tb \leq \|a \|^2 + \|b\|_*^2$ with dual norm
Aug
18
comment Closed forms for two times series similar to geometric series, but with additional power
Thanks for the answer! What I am interested in is the growth rate of the series with respect to the upper bound of the summation, which is $T$.
May
31
comment Hölder's inequality/Cauchy-Schwarz for Bregman Divergence?
Good point. If F is strongly convex (equivalently F* is smooth) we should be able to get it. But still have not written it in a clean form.
May
30
revised Hölder's inequality/Cauchy-Schwarz for Bregman Divergence?
added 89 characters in body
May
30
revised Hölder's inequality/Cauchy-Schwarz for Bregman Divergence?
added 89 characters in body
May
29
asked Hölder's inequality/Cauchy-Schwarz for Bregman Divergence?
May
25
asked Deriving Dual Averaging from (Sub)gradient Descent
May
18
revised Smooth approximation of maximum using softmax?
edited body
May
17
revised Closed form for sequence: $\sum_{j=1}^k 2^j j^{1/2}$
added 4 characters in body
May
17
asked Closed form for sequence: $\sum_{j=1}^k 2^j j^{1/2}$
May
5
revised Finding $Q$ for any $A$ s.t. $QAQ^\top = I$
added 8 characters in body
May
5
comment Finding $Q$ for any $A$ s.t. $QAQ^\top = I$
@AlgebraicPavel I literally do whatever you just suggested. But is it the best way we can do this? I want to get rid of the inversion D^-0.5