Lecture5-1Recap and Preview

Flow Chart Recap

2 Central Questions

Learning is split into 2 questions

- Can we make sure that Ein≈Eout?

- Can we make Ein small enough?

Trade off on M(# of hypothesis set)

Small M :) few choices :(

Large M :( many choices :)

Using the right M
4000
is important! M=∞ is doomed?

Preview

Known

ℙ[|Ein(g)−Eout(g)|≥ϵ]≤2Mexp(−2ϵ2N)

To do

establish a finite quantity m that replaces M

ℙ[|Ein(g)−Eout(g)|≥ϵ] ? 2mexp(−2ϵ2N)

Justify the feasibility of learning for infinite M

study m to understand its trade-off for ‘right’ , just like M