Uncertainty-based Deep Learning Networks for Limited Data Wetland User Models

EasyChair Preprint 675

Abstract

A method for dealing with limited data in deep networks is given which is based upon calculating uncertainty associated with remaining data. The method was developed for the Watershed REstoration using Spatio-Temporal Optimization of REsources (WRESTORE) system, an interactive decision support system designed for performing multi-criteria decision analysis with a distributed system of conservation practices on the Eagle Creek Watershed. Eventually, these results from neural network user models may be integrated in to an existing genetic algorithm to find Pareto optimal solutions for multiple stakeholders within the constraints of the physical and socio-economic environment. Our results show faster and more stable convergence when using an uncertainty based incremental sampling method than when using a standard random incremental sampling method. This work describes the existing WRESTORE system, provides details about the implementation of our uncertainty based incr! emental sampling method, and provides discussion of our results and future work.

Keyphrases: Deep Neural Network, Hydrology, Interactive Genetic Algorithm, Interactive Machine Learning, deep learning, incremental learning, incremental sampling method, interactive decision support system, interactive optimization, limited data, multi-objective optimization, uncertainty, uncertainty based incremental sampling, uncertainty based sampling method

@booklet{EasyChair:675,
  author    = {Andrew Hoblitzell},
  title     = {Uncertainty-based Deep Learning Networks for Limited Data Wetland User Models},
  doi       = {10.29007/vhf2},
  howpublished = {EasyChair Preprint 675},
  year      = {EasyChair, 2018}}