Current project entering final validation phase

Stay tuned:

https://www.mql5.com/de/users/dasunding/seller

Rapid convergence: Within 140 minutes, the agent reached a profit factor of >1.2, successfully demonstrating the viability of the integrated RL framework and feature set in out of sample data.

Enhanced regularization and gradient accumulation led to a substantial increase in model performance. Further testing is required.

ARTEMIS Alpha AI - Reinforced Multiscale Temporal Fusion Transformer Status:

Successful validation of initial ONNX inference within MQL5

Improving pretrained DQN-Agent

Milestone: MQL5 performance has achieved approximately 98% parity with the Python simulation results. This represents a highly acceptable margin, especially considering that the Python model utilizes 30-minute bar data, whereas the MQL5 environment operates on granular tick data. Performance metrics remain model-specific and are subject to variation across different neural network iterations.

For optimal results with the Portfolio Optimizer, it's recommend selecting markets with minimal correlation. This ensures that the generated machine learning strategies are more likely to be uncorrelated, enabling higher exposure without increasing risk.

Whitepaper Update (Ger)