1: \begin{abstract}

2: Weak gravitational lensing is a powerful cosmological probe, with

3: non--Gaussian features potentially containing the majority of the

4: information. We examine constraints on the parameter triplet

5: $(\Omega_m,w,\sigma_8)$ from non-Gaussian features of the weak lensing

6: convergence field, including a set of moments (up to $4^{\rm th}$

7: order) and Minkowski functionals, using publicly available data from

8: the 154\,deg$^2$ CFHTLenS survey. We utilize a suite of ray--tracing

9: N-body simulations spanning 91 points in $(\Omega_m,w,\sigma_8)$

10: parameter space, replicating the galaxy sky positions, redshifts and

11: shape noise in the CFHTLenS catalogs. We then build an emulator that

12: interpolates the simulated descriptors as a function of

13: $(\Omega_m,w,\sigma_8)$, and use it to compute the likelihood function

14: and parameter constraints. We employ a principal component analysis to

15: reduce dimensionality and to help stabilize the constraints with

16: respect to the number of bins used to construct each statistic. Using

17: the full set of statistics, we find

18: $\Sigma_8\equiv\sigma_8(\Omega_m/0.27)^{0.55}=0.75\pm0.04$ (68\%

19: C.L.), in agreement with previous values. We find that constraints on

20: the $(\Omega_m,\sigma_8)$ doublet from the Minkowski functionals

21: suffer a strong bias. However, high-order moments break the

22: $(\Omega_m,\sigma_8)$ degeneracy and provide a tight constraint on

23: these parameters with no apparent bias.  The main contribution comes

24: from quartic moments of derivatives.

25: \end{abstract}

26: