Supervised learning assumes underlying structure

Supervised learning assumes that outputs are a function of inputs combined with noise. A “good model” estimates the original function with high accuracy. Loss formalizes this mismatch.

Resources

Backlinks