Starting from a simple deterministic model, we show that the asymptotic outcomes (as time goes to infinity) of both shallow and deep neural networks such as those used in BloombergGPT to generate economic time series are exactly the Nash equilibria of a non-potential game. We then analyse deep neural network algorithms that converge to these equilibria. The approach is extended to federated deep neural networks between clusters of regional servers and on-device clients. Finally, the variational inequalities behind large language models including encoder-decoder related transformers are established.
This video was produced by the SITE Research Center at New York University, as part of their talk series.
