/gradient-descent-custom-loss

Primary LanguageJupyter Notebook

Implementing Linear and Polynomial regression with custom loss metrics.

Linear regression had to be implemented with the loss functions :

a) |x -  x̂|³
b) |x -  x̂|

Polynomial regression(Degree=2) had to be implemented with the loss functions :

a) |x -  x̂|⁴
b) |x -  x̂|⁷

The higher power loss functions were a bit difficult to work with, becuase they were exploding too quickly. The gradients and the loss both had to be controlled so that they don't overflow the limits.