1: \begin{abstract}

2: We devise a new algorithm, called the \ree{fast} linear response algorithm, for accurately differentiating SRB measures with respect to some parameters of the dynamical system, where SRB measures are \ree{fractal} limiting stationary measures of chaotic systems. 

3: 

4: The core of our algorithm is the first numerical treatment of the unstable divergence, 

5: a central object in the linear response theory for \ree{fractal attractors}. 

6: We derive the first computable expansion formula of the unstable divergence,

7: where all terms are functions rather than distributions. 

8: Then we give a `fast' characterization of the expansion

9: by renormalized second-order tangent equations, 

10: whose second derivative is taken in a modified shadowing direction, 

11: computed by the non-intrusive shadowing algorithm.

12: 

13: The new characterization makes the algorithm efficient and robust: 

14: its main cost is solving $u$, the unstable dimension,

15: many first-order and second-order tangent equations, 

16: and it does not compute oblique projections. 

17: Moreover, the algorithm works for chaos on Riemannian manifolds with any $u$;

18: its convergence to the true derivative is proved for uniform hyperbolic systems.

19: The algorithm is illustrated on an example which is difficult for previous methods.

20: 

21: \smallskip

22: \noindent \textbf{Keywords.}

23: Linear response, fast algorithm, SRB measure, chaos, non-intrusive shadowing.

24: \end{abstract}

25: