PAPER / ARXIV:2608.23393
Li, J.
RESUMO
KellyBoost is a single multi-output XGBoost model whose softmax output is the portfolio: with y vector of per-asset holding-period returns, training loss is -log(1 + w·y), negative log growth rate, so the fitted model is growth-optimal (Kelly) allocation conditioned on the features. The objective is exact rather than surrogate: we derive the gradient, analytic diagonal Hessian and full Hessian in closed form, verify them by finite differences, and ship a dependency-free reference engine.
NO MESMO MAPA