Discussing the article: "Neural Networks in Trading: Contrastive Pattern Transformer"
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Check out the new article: Neural Networks in Trading: Contrastive Pattern Transformer.
When analyzing market situations using machine learning, we often focus on individual candlesticks and their attributes, overlooking candlestick patterns that frequently provide more meaningful information. Patterns represent stable candlestick structures that emerge under similar market conditions and can reveal critical behavioral trends.
Previously, we explored the Molformer framework, borrowed from the domain of molecular property prediction. The authors of Molformer combined atomic and motif representations into a single sequence, enabling the model to access structural information about the analyzed data. However, this approach introduced the complex challenge of separating dependencies between nodes of different types. Fortunately, alternative methods have been proposed that avoid this issue.
One such example is the Atom-Motif Contrastive Transformer (AMCT), introduced in the paper "Atom-Motif Contrastive Transformer for Molecular Property Prediction". To integrate the two levels of interactions and enhance the representational capacity of molecules, the authors of AMCT proposed to contrastive learning between atom and motif representations. Since atom and motif representations of a molecule are essentially two different views of the same entity, they naturally align during training. This alignment allows them to provide mutual self-supervised signals, thereby improving the robustness of the learned molecular representations.
Author: Dmitriy Gizlyk