The True Reason Why We Need Regularization

Regularize your way to a better tomorrow

Data science depends on math, and math is generally focused on situations where:

But they do matter, as there can be exceptions:

Problems suffering from the above are called ill-posed problems. Relating to linear algebra and systems of equations, the only truly well-posed problems are those with a single unique solution.

A lot of what we covered with linear regression was about getting matrices into the right shape for them to be solvable in this sense. But some matrices just won’t submit to this, and other problems may technically “fit” linear regression but still be violating the above assumptions in subtle ways.

How is this related to well and ill-posed problems? It’s not clearly a no solution or multiple solution case, but it does fall in the third category — overfitting results in fitting to the…

Related posts:

Did you know the majority of us live in our head but there is so much more wisdom in our body it has been said. So settle in and take a pause, let’s calm the nervous system and explore. So set the…

Mental Illness has only recently become an “acceptable” topic in mainstream society. More people are being diagnosed, and seeking help, than ever before. Celebrities and influencers are speaking out…

The After Action Review (AAR) is a short, sharp process to improve an organisation ability to learn in the midst of action and improve team working and knowledge management. A single AAR can have…