I've considered gradient descent for optimizing parameters on toy problems at university a few times. Never actually did it though, it's a lot of hassle for the

I've considered gradient descent for optimizing parameters on toy problems at university a few times. Never actually did it though, it's a lot of hassle for the advantage of less interaction at the cost of no longer building some intuition.