1: \begin{abstract}

2: %We consider solving optimizations with differentiable objective function constrained to an arbitrary (possibly non-convex) set.

3: %We show that the simple iterative projected gradient (IPG) algorithm can achieve a near-global-optimum solution. 

4: We study convergence of the iterative projected gradient (IPG) algorithm for arbitrary (possibly non-convex) sets and when both the gradient and projection oracles are computed approximately. We consider different notions of  approximation of which we show that 

5: %three classes of approximate oracles: fixed-precision,

6: the Progressive Fixed Precision (PFP) and the $(1+\epsilon)$-optimal oracles can achieve the same  accuracy as for the exact IPG algorithm. We show that the former scheme is also able to maintain the (linear) rate of convergence of the exact algorithm, under the same embedding assumption. In contrast, the $(1+\epsilon)$-approximate oracle requires a stronger embedding condition, moderate compression ratios and it typically slows down the convergence. 

7: 

8: We apply our results to accelerate solving a class of data driven compressed sensing problems, where we replace iterative 

9: %Finally we study an application of the inexact IPG in a data driven compressed sensing problem. We shortcut the computations involved with (iteratively) 

10:  exhaustive searches over large datasets by fast approximate nearest neighbour search strategies based on the cover tree data structure. For datasets with low intrinsic dimensions our proposed algorithm achieves a complexity logarithmic  in terms of the dataset population as opposed to the linear complexity of a brute force search. 

11: By running several numerical experiments we conclude similar observations as predicted by our theoretical analysis.

12: 

13: 

14: % approximation errors Under certain assumptions on 

15: \end{abstract}