Learning rate

As we prefer a gradual decrease in learning rate from a high value to low value during training, select the best explanation which supports this hypothesis.

If the decay rate is slow, a lot of time will be wasted bouncing around with little improvement in the loss. If the decay rate is too high, the learning rate will decay soon to very less value and unable to reach best minima.

If the decay rate is slow, it will explore the dataset with significant improvement in the loss. If the decay rate is too high, the learning rate will decay soon to very less value and unable to reach best minima.

If the decay rate is slow, it will explore the dataset with significant improvement in the loss. If the decay rate is too high, the learning rate will high as$,$ after initial exploration, it will find best minima in a short amount of time.

If the decay rate is slow, a lot of time will be wasted bouncing around with little improvement in the loss. If the decay rate is too high, the learning rate will high as$,$ after initial exploration, it will find best minima in a short amount of time.