Differentiable Ghost Operators: A Physics-Informed Neural Turing Machine for Cross-Sectional Stock Selection

Extracting Alpha in extreme low signal-to-noise ratio (SNR) environments, such as the Chinese A share market, remains a notoriously unsolved challenge for deep learning. Traditional heavily parameterized models, including Transformers, inevitably fall into the ”dimensionality disaster,” memorizing market noise rather than fundamental mechanics. To break this overfitting curse, we propose a novel, ultra-lightweight architecture inspired by thermody namics and Neural Turing Machines (NTMs): the Physics-Informed Ghost Operator. By mapping the cross-sectional stock market into a high-dimensional physical manifold, our Ghost Operators navigate the feature space driven by gravitational routing. Crucially, we enforce a minimal action principle via a Boltzmann-distributed temperature scaling and Pauli-exclusion-like potential well clamping. Walk forward validation on 10 years of real A-share data reveals that our architecture achieves a substantial Sharpe ratio improvement (up to 3.2x) and cuts the maximum drawdown by nearly half compared to native NTMs. Furthermore, network sparsity is reduced by 66%, proving that physical constraints compel the model to aggressively filter noise and focus strictly on high-potential Alpha regions.