Title: Inequalities for sums of random variables in noncommutative probability spaces

Abstract: We extend the classical Hoeffding inequality to the noncommutative setting. Under certain conditions we establish an improved version of the noncommutative Bennett inequality and use it to prove an extension of Rosenthal inequality as follows: Let $(\mathfrak{M}, \tau)$ be a noncommutative probability space, $\mathfrak{N}$ be a von Neumann subalgebra of $\mathfrak{M}$ with the corresponding conditional expectation $\mathcal{E}_{\mathfrak{N}}$ and let subalgebras $\mathfrak{N}\subseteq\mathfrak{A}_j\subseteq\mathfrak{M}\,\,(j=1, \cdots, n)$ be successively independent over $\mathfrak{N}$. If $1\leq r\leq2\leq p<\infty$ and $x_j\in\mathfrak{A}_j$ are self-adjoint elements such that $\mathcal{E}_{\mathfrak{N}}(x_j)=0$ for all $1\leq j\leq n$, then \begin{eqnarray*} \left\|\sum_{j=1}^nx_j\right\|_p^p\leq C(p, r)\left\{\sum_{j=1}^n\|x_j\|_p^p+\left(\sum_{j=1}^n\|x_j^r\|\right)^{\frac{p}{r}}\right\}. \end{eqnarray*}