Update History

1. [2023.10.28] Problem 4 (1): correct Eq. (4.3):

   $$\begin{array}{}& {\mathcal{D}}_{\psi }({\mathit{p}}_t,{\tilde{\mathit{p}}}_{t+1})=\sum _{i=1}^N{c}_i\cdot p_{t,i}(\eta \cdot \frac{{\ell }_{t,i}}{c_i}-1+\exp (-\eta \cdot \frac{{\ell }_{t,i}}{c_i})).\end{array}$$

2. [2023.10.29] AOpt-Lab1 score update: - 40/40: Tune the parameters so that by the 20th round, the loss outputted by the FISTA algorithm should be < 23,000.

3. [2023.10.31] Problem 5 (3): Mirror Descent Lemma refine:

   $$\begin{array}{}& f_t\left({\mathbf{x}}_t\right)-f_t(\mathbf{u})\le \frac{{\eta }_t}{\lambda }{\Vert \mathrm{\nabla }{f}_t\left({\mathbf{x}}_t\right)\Vert }_{\star }^2+\frac{1}{{\eta }_t}({\mathcal{D}}_{\psi }(\mathbf{u},{\mathbf{x}}_t)-{\mathcal{D}}_{\psi }(\mathbf{u},{\mathbf{y}}_{t+1})-{\mathcal{D}}_{\psi }({\mathbf{y}}_{t+1},{\mathbf{x}}_t)).\end{array}$$

4. [2023.11.07] AOpt-Lab1: don't need to calculate the smoothness parameter Please fill in the content for TODO in the code block below to calculate the Smoothness parameter $L$$L$, just tune the $30<L<300$$30<L<300$.