Quantum-Inspired Neural Networks with Markov States and Backpropagation

Summary

The article describes an MQL5 trading model that combines neural-network training with Markov market-state labels, a 256-dimensional state-space model, and mechanisms named resonance, interference, and decoherence. It describes inputs spanning prices, volume, technical indicators, candlestick patterns, and time features. Rare-class weighting is applied to the loss so that strong rises and falls receive more attention than common flat periods; Adam is used for training.

The article reports accuracy of up to 65%, drawdown no greater than 20%, and Sharpe ratios between 1.8 and 2.4. It also claims that adding the state-space component improved accuracy by 7% over discrete states alone, and gives class-wise results after weighting. These are author-reported experiments; the excerpt does not provide enough detail about data, assets, sample periods, validation design, costs, or out-of-sample testing to assess robustness. The quantum terminology describes mathematical analogies in a conventional software model, not demonstrated quantum computing.

Key ideas

• The model combines five discrete market regimes with a hidden state intended to retain longer-term quantitative information.
• Resonance, interference, and decoherence are implemented as mathematical transformations of inputs and context.
• Inverse-frequency class weights increase the loss contribution of underrepresented market classes.
• The article reports performance figures, but provides limited information for independently evaluating their reliability.
• The quantum language is analogical; the implementation described is a conventional neural network.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.