machine learning - Why do we minimise a cost function instead of maximising an equivalent? - Cross Validated