Loss convergence in a causal Bayesian neural network of retail firm\n performance

2020/08/29 by F. Trevor Rogers, Rogers, F. Trevor
Engineering · Decision Sciences · #Energy Load and Power Forecasting #Stock Market Forecasting Methods #Auction Theory and Applications

paper · pdf · doi:10.48550/arxiv.2008.13038

Abstract

We extend the empirical results from the structural equation model (SEM)\npublished in the paper Assortment Planning for Retail Buying, Retail Store\nOperations, and Firm Performance [1] by implementing the directed acyclic graph\nas a causal Bayesian neural network. Neural network convergence is shown to\nimprove with the removal of the node with the weakest SEM path when variational\ninference is provided by perturbing weights with Flipout layers, while results\nfrom perturbing weights at the output with the Vadam optimizer are\ninconclusive.\n

Related

• Concurrent Neural Network : A model of competition between times series
• Deep Learning for Limit Order Books
• ARMDN: Associative and Recurrent Mixture Density Networks for eRetail Demand Forecasting
• A Dynamic Bayesian Network Model for Inventory Level Estimation in Retail Marketing
• The Causal Loss: Driving Correlation to Imply Causation