Hoeffding inequality for bounded independent random variables

Let $X_1$, $X_2,\mathrm{\dots },X_n$ be independent random variables, such that $\mathrm{Pr}(a_k\le X_k\le b_k)=1$ for all $k$, where $a_k$ and $b_k$ are constant, . Let $S_n$ be the sum $X_1+\mathrm{\dots }+X_n$. Then

$$\mathrm{Pr}(S_n-E[S_n]>ϵ)\le \exp \left(-\frac{2{ϵ}^2}{{\sum }_{k=1}^n{(b_k-a_k)}^2}\right),$$

$$\mathrm{Pr}(|S_n-E[S_n]|>ϵ)\le 2\exp \left(-\frac{2{ϵ}^2}{{\sum }_{k=1}^n{(b_k-a_k)}^2}\right).$$