3.2) Which of the following statements is/are true?

Generally, while minimizing the loss, local minima is guaranteed but not the global

minimum.

In a high-dimensional space, saddle points are quite common and hence advanced

optimization techniques, such as momentum-based methods can be employed.

It is difficult to come out of the saddle points, because of the plateau or flat regions,

which gives an impression that the gradient is zero across dimensions.