(Solved): Gradient Descent Optimization (a) Consider Figure 5, which depicts a loss function L(x):R2R. The ...

Gradient Descent Optimization (a) Consider Figure 5, which depicts a loss function $L(x):{\mathbb{R}}^2?\mathbb{R}$. The red dot represents the current estimate of ${\mathbf{x}}_t=\left[x_1,x_2\right]$ at step $t$. Please sketch an estimate of the path of 2 updates that would be taken by vanilla SGD until "convergence". Hint: use the tangent vector to help illustrate your reasoning. Figure 5: Contour lines of an arbitrary cost function with current estimate ${\mathbf{x}}_t$.