Latest

This article investigates the performance discrepancy between Autoformer and DLinear models as reported in the sources. It attempts to replicate the findings of the original paper and the blog post, and we discuss the potential differences in methods and preprocessing between the two sources. It also presents its own experiments and compare the two models across different datasets and variations. The results show that DLinear consistently outperformed Autoformer in our tests. It concludes by reflecting on the potential impact of the research on the understanding and application of transformer-based models in time-series analysis.

Understanding the basics of Neural Networks