Relation to Concept Learning

In concept learning we have an instance space $X$, hypothesis space $H$, training examples $D$. The FindS algorithm finds the most specific hypothesis from $VS_{H,D}$.
Assume fixed set of instances $\langle x_{1}, \ldots, x_{m}\rangle$
Assume $D$ is the set of classifications $D = \langle c(x_{1}), \ldots, c(x_{m})\rangle$
Choose $P(D\,|\,h)$ \[ P(D\,|\,h) = { \{ \array{ 1 & if \forall_{d_i \in D} d_i = h(x_i)\\ 0 & otherwise } } \]
Choose $P(h)$ to be uniform distribution so that $P(h) = \frac{1}{|H|}$ for all $h$ in $H$
Then we can say that \[ P(h\,|\,D) = { \{ \array{ \frac{1}{|VS_{H,D}|} & if h is consistend with D \\ 0 & otherwise. } } \]
What happens in Brute force BCL is that all hypotheses start with with same probability then the inconsistent hypotheses drop to zero while the rest share equally.
Every consistent hypothesis is a MAP hypothesis.