This episode explores the application of incremental learning, specifically using a Stochastic Gradient Descent (SGD) model, for building a robot trading system. It contrasts this approach with batch-supervised learning and reinforcement learning, arguing that incremental learning is more suitable for the non-stationary and rapidly changing nature of financial data.
The article presents a simplified demonstration of how such a system can be built and trained, while also cautioning against over-optimistic expectations of guaranteed profit.
This audio podcast is produced by the author using generative artificial intelligence technology, and the article describing the end-to-end process of developing the nowcasting model is used as the main source.
