1: \begin{abstract}

2: We study approximations of non-Gaussian stationary processes 

3: having long range correlations with microcanonical models. 

4: These models are conditioned by the empirical value of an

5: energy vector, evaluated on a single realization.

6: Asymptotic properties of maximum entropy microcanonical 

7: and macrocanonical processes and their convergence to Gibbs measures are reviewed. We show that 

8: the Jacobian of the energy vector controls the entropy rate of microcanonical processes. 

9: 

10: Sampling maximum entropy processes through MCMC algorithms

11: require too many operations when the number of constraints is large.

12: We define microcanonical gradient descent processes by

13: transporting a maximum entropy measure with a gradient descent algorithm

14: which enforces the energy conditions. Convergence and symmetries are

15: analyzed. 

16: Approximations of non-Gaussian processes with long range interactions

17: are defined with

18: multiscale energy vectors computed with wavelet and scattering transforms.

19: Sparsity properties are captured with $\bf l^1$ norms. 

20: Approximations of Gaussian, Ising and point processes are studied, as well

21: as image and audio texture synthesis.

22: \end{abstract}

23: