The concept of probably approximately correct (PAC) learning has been foundational in computational learning theory.
To solve a problem, you have to be willing to think differently.
The more perfect the artist, the more completely separate in him will be the man who suffers and the mind which creates.
Every failure is a step closer to the answer.
A problem is not solved by mechanically applying past experience, but by restructuring the situation.