Source: Deep Learning on Medium

Regularization in machine learning, How to tune a man into Marriage material

High Variance

Okay, you want to tune a man into husband for production, you are an ML engineer and you know how to tune a model to decrease generalization error.

As you are happy with the religious aspect of life and you want to the same to be in him, so you fed him with all types of religions on earth and their hidden values and you are surprised to see a monk ready to teach you the state of art religious learning techniques

Reverted the training process, and you want to tune the Laziness dimension of him …. so you taught him to kill the procrastination with brain beats, surprised to see again as he was digging a tunnel for the apocalypse for about 2KM down your house and complaining about your procrastination.

You send him Deeplearning school ,Is your Man writing an LSTM Rnn to generate a calculator in python or simple if-else on the math package? which has more accuracy !!!

TOOO Much , Overfit , learnt beyond, memorised everything , inherited the statistical noise……. all are the same for high variance.

High bias

He had an iq of around 11.40 and some how you made him a nuclear mission head of your father SpaceX, what does he do as cheif ! using some radium mineral to paint stars on the house roof for romantic dinner ?

trying share market shots with huge investment’s with your money,can you accept it or do you invest to see him to act in a movie with a million dollar budget .

Underfit, ameteur, partial learning , shallow knowledge ,under trained ,loosely coupled ……. all are the same for high bias.