Loss convergence in a causal Bayesian neural network of retail firm performance
This work addresses convergence issues in causal Bayesian neural networks for retail performance modeling, but it appears incremental as it builds directly on prior structural equation model research.
The authors tackled the problem of improving neural network convergence in a causal Bayesian neural network modeling retail firm performance by removing the node with the weakest structural equation model path, showing improved convergence with Flipout layers but inconclusive results with the Vadam optimizer.
We extend the empirical results from the structural equation model (SEM) published in the paper Assortment Planning for Retail Buying, Retail Store Operations, and Firm Performance [1] by implementing the directed acyclic graph as a causal Bayesian neural network. Neural network convergence is shown to improve with the removal of the node with the weakest SEM path when variational inference is provided by perturbing weights with Flipout layers, while results from perturbing weights at the output with the Vadam optimizer are inconclusive.