1 / 3 Begin with the evidence
Six points from the past, nothing from tomorrow
These points are everything the model has seen. We have not yet chosen the rule that produced them.
training loss · simple rule—
training loss · alternative rule—
prediction at x = 7hidden
simple rule
another rule matching the same past
first hypothesis
h₁(x) = 2x + 1second hypothesis
h₂(x) = 2x + 1 + α∏⁶ᵢ₌₁(x − i)The past gave you points. Calling them a pattern still adds an assumption.