The following is from Understanding Machine Learning: Theory to Algorithm textbook:

Definition of PAC Learnability: A hypothesis class $\mathcal H$ is PAC learnable if there exist a function $m_H : (0, 1)^2 \rightarrow \mathbb{N}$ and a learning algorithm with the following property: For every $\epsilon, \delta \in (0, 1)$, for every distribution $D$ over $X$, and for every labeling function $f : X \rightarrow \{0,1\}$, if the realizable assumption holds with respect to $\mathcal H,D,f$ then when running the learning algorithm on $m \ge m_H(\epsilon,\delta)$ i.i.d. examples generated by $D$ and labeled by $f$, the algorithm returns a hypothesis $h$ such that, with probability of at least $1 - \delta$ (over the choice of the examples), $L_{(D,f)}(h) \le \epsilon$.

1) In the function definition $m_H : (0, 1)^2 \rightarrow \mathbb{N}$; what does a) 0 and 1 in the bracket, b) the integer 2, and c) $\rightarrow \mathbb{N}$ refer to?


The explanations are as follows:

  • $m_H:(0,1)^2 \rightarrow \mathbb N$ is a similar notation to $f:R^n\rightarrow \mathbb N$ which means it takes a n-dimensional input consisting of real numbers only. In the case of PAC learning the input is 2 dimensional consisting of numbers between $0$ and $1$ only which stand for values of $\epsilon, \delta$ respectively.

  • The integer $2$ as explained above is the dimension of the input vector.

  • $\rightarrow \mathbb N$ means mapped to natural numbers. In case of PAC learning that for each value of $(\epsilon, \delta)$ a function $m_H$ maps it to natural numbers, simply put $m_H(\epsilon, \delta) = $

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.