1: \begin{abstract}

2:   Existing adaptive bias techniques, which seek to estimate free

3:   energies and physical properties from molecular simulations, are

4:   limited by their reliance on fixed kernels or basis sets which 

5:   hinder their ability to efficiently conform to varied free energy 

6:   landscapes.  Further, user-specified parameters are in general 

7:   non-intuitive, yet significantly affect the convergence rate and 

8:   accuracy of the free energy estimate.  Here we propose a novel method 

9:   wherein artificial neural networks (ANNs) are used to develop an 

10:   adaptive biasing potential which learns free energy landscapes. 

11:   We demonstrate that this method is capable of rapidly adapting to 

12:   complex free energy landscapes and is not prone to boundary or oscillation 

13:   problems. The method is made robust to hyperparameters and overfitting 

14:   through Bayesian regularization which penalizes network weights and 

15:   auto-regulates the number of effective parameters in the network.  

16:   ANN sampling represents a promising innovative approach which can 

17:   resolve complex free energy landscapes in less time than conventional 

18:   approaches while requiring minimal user input.

19: \end{abstract}

20: