1: \begin{abstract}

2: 

3:   \noindent In this paper we develop proximal methods for statistical learning.

4:   Proximal point algorithms are useful in statistics and

5:   machine learning for obtaining optimization solutions for composite functions.

6:   Our approach exploits closed-form solutions of proximal operators and

7:   envelope representations based on the Moreau,

8:   Forward-Backward, Douglas-Rachford and Half-Quadratic envelopes. Envelope

9:   representations lead to novel proximal algorithms for statistical

10:   optimisation of composite objective functions which include both non-smooth and

11:   non-convex objectives.  We illustrate our methodology with

12:   regularized Logistic and Poisson regression and non-convex bridge penalties

13:   with a fused lasso norm.  We provide a discussion of convergence of

14:   non-descent algorithms with acceleration and for non-convex functions.  

15:   Finally, we provide directions for future research.

16: 

17:   \vspace{0.1in}

18:   \noindent Keywords: Bayes MAP; shrinkage; sparsity; splitting; fused lasso;

19:   Kurdyka-\L ojasiewicz; non-convex

20:   optimisation; proximal operators; envelopes; regularization;  ADMM;

21:   optimization; Divide and Concur. 

22: \end{abstract}

23: