New answers tagged

2 votes
Accepted

Learning with zero inductive bias

To be more precise, if you want a distribution-free generalization bound, then you must have some inductive bias (these are the no-free-lunch theorems referenced by D.W.). For binary classification, ...
Aryeh's user avatar
  • 10.3k
1 vote

Learning with zero inductive bias

You cannot. See the No free lunch theorem (e.g., here and here and here and many other resources).
D.W.'s user avatar
  • 11.7k
1 vote

Learning positive half-lines (in $\mathbb{N}$)

Yes, it's trivial. The learning algorithm consists of choosing the smallest positive example $x_0$, and taking the hypothesis to be $h(x)=1[x\ge x_0]$. All of the generalization guarantees proved for $...
Aryeh's user avatar
  • 10.3k
1 vote
Accepted

Learning arithmetic series

An arithmetic series is defined by the 1st term $t_1$ and the difference between terms $d$. If you stipulate that $\max(|t_1|,d)\le M$ then you have a finite hypothesis space and hence a finite ...
Aryeh's user avatar
  • 10.3k
0 votes

Why is the estimation error smaller in Structural Risk Minimization

The answer is right under my noise but I'd failed to see it. Theorem 7.4 in the book says that: $$ \mathbb{P}\lbrace S\sim\mathcal{D}^m: (\forall h\in\mathcal{H}) (L_\mathcal{D}(h)\le L_S(h) + \...
Tran Khanh's user avatar
0 votes
Accepted

Unable to understand the Sample complexity of PAC learning

I don't understand exactly your question, but I'll answer it from the two possible misunderstandings I can see. The first confusion comes from your definition of the function $m_\mathcal{H} : (0, 1)^2 ...
Ayoubayjx's user avatar
  • 122

Top 50 recent answers are included