2020/09/16 by Paul A. Parker, Parker, Paul A., Scott H. Holan +1
Computer Science · #Machine Learning and ELM #Gaussian Processes and Bayesian Inference
paper · pdf · doi:10.48550/arxiv.2009.07934
The topic of deep learning has seen a surge of interest in recent years both\nwithin and outside of the field of Statistics. Deep models leverage both\nnonlinearity and interaction effects to provide superior predictions in many\ncases when compared to linear or generalized linear models. However, one of the\nmain challenges with deep modeling approaches is quantification of uncertainty.\nThe use of random weight models, such as the popularized "Extreme Learning\nMachine," offer a potential solution in this regard. In addition to uncertainty\nquantification, these models are extremely computationally efficient as they do\nnot require optimization through stochastic gradient descent, which is what is\ntypically done for deep learning. We show how the use of random weights in a\ndeep model can fit into a likelihood based framework to allow for uncertainty\nquantification of the model parameters and any desired estimates. Furthermore,\nwe show how this approach can be used to account for informative sampling of\nsurvey data through the use of a pseudo-likelihood. We illustrate the\neffectiveness of this methodology through simulation and with a real survey\ndata application involving American National Election Studies data.\n