4 ms·
550–600 million years of hyperparameter tuning and neural architecture search using an evolutionary algorithm is nothing to laugh at. But the actual learning
by MAXPOOL 6y ago
550–600 million years of hyperparameter tuning and neural architecture search using an evolutionary algorithm is nothing to laugh at.
But the actual learning process of a single brain uses very little training examples and is very energy efficient. A brain that uses only 20 watts of power.
- polishdude20 6y agoYeah it's as if we evolved to have a brain that is good at learning itself. We aren't good at any one thing except learning.