1: \begin{abstract}

2: In this work, we analyze the regularizing property of the stochastic gradient descent

3: for the efficient numerical solution

4: of a class of nonlinear ill-posed inverse problems in Hilbert spaces. At each step of the iteration, the method randomly

5: chooses one equation from the nonlinear system to obtain an unbiased stochastic estimate of the gradient, and then performs a

6: descent step with the estimated gradient. It is a randomized version of the classical Landweber method for nonlinear

7: inverse problems, and it is highly scalable to the problem size and holds significant potentials for solving large-scale inverse problems.

8: Under the canonical tangential cone condition, we prove the regularizing property for \textit{a priori} stopping rules,

9: and then establish the convergence rates under suitable sourcewise condition and range invariance condition.

10: 

11: \smallskip

12: \noindent{\bf Keywords}:

13: stochastic gradient descent, regularizing property, nonlinear inverse problems, convergence rates

14: 

15: \smallskip

16: \noindent{\bf AMS Subject Classifications}: 65J20, 65J22, 47J06

17: %\fn{Zehui, please read some papers that discuss SGD and

18: %find the AMS Subject Classifications used there. Then list each of them here, along with the actual meanings

19: %of these classifications that you can find from the online AMS Subject Classifications.

20: %Then Bangti may help find 3 or 4 appropriate AMS Subject Classifications.}

21: \end{abstract}