In briefAuction-based allocation of routes in capacity-constrained networks
SummaryThis paper studies how to allocate sets of routes between end nodes in capacity-constrained flow networks with transfer costs. The authors introduce an origin-destination bidding language, where participants bid on bounded conic combinations of inlet, outlet and quantity licences, and routing is determined by a network-flow winner-determination problem. Under a restricted model, the VCG payment rule is shown to be core-stable. For the general model, where source access may depend on the individual consumer, they propose a nodal pricing rule based on shadow prices of flow-conservation constraints, which is always core-stable but not dominant-strategy incentive compatible.
In briefData-driven dynamic portfolio selection using vine copula simulation and reinforcement learning
SummaryThis paper proposes a dynamic portfolio selection method that combines vine copula simulation with reinforcement learning, for an investor with constant relative risk aversion who penalises conditional value-at-risk (CVaR) of losses. A conditional vine model is fitted to historical returns to generate scenarios and paths for training a recurrent portfolio policy, and stability bounds are derived for parameter errors. On a seven-asset panel of US and Chinese equities and gold, an ensemble of ten trained policies reports a 34.13% compound annual growth rate and a Sharpe ratio of 2.31 over 22 monthly holding periods, exceeding six benchmarks.
In briefStructured neural solver for option pricing under stochastic volatility
SummaryThis paper introduces BeTINN, a structured conditional neural architecture that extends time-induced modulation with multi-directional time-parameter conditioning and a joint residual pathway. It is evaluated on the Heston stochastic-volatility PDE, improving joint accuracy for normalized option value and key sensitivity metrics over the authors' baselines. The authors also find that validation physics loss does not fully determine derivative fidelity, so derivative-aware evaluation is recommended when sensitivities matter.
In briefThe Missile and the Toll: Two-Way Flow Through the Strait of Hormuz
SummaryThis paper models a Stackelberg escalation game in which an enforcer picks an attack intensity and a toll schedule for the Strait of Hormuz, while ships choose to pay, cross exposed, wait, reroute, or exit. The authors argue that attacks create willingness to pay and tolls capture it. The model predicts asymmetry across trade types. Payments would be clandestine, but its implications can be tested against freight rates, insurance premia, and incident reports.
In briefDynamic industry-network survival model for mortgage default risk
SummaryThis paper proposes a discrete-time survival model that measures economic proximity by industrial composition similarity and rebuilds dynamic industrial graphs annually to capture changing cross-area dependence in mortgage default risk. It combines a conditional autoregressive process with a first-order autoregressive process and estimates the model using Integrated Nested Laplace Approximation. Tested on U.S. mortgage loan-level data, the dynamic model outperforms static industrial graph specifications in in-sample fit and out-of-sample discrimination.
In briefA GLE-LSV model with state-dependent leverage and colored noise
SummaryThis paper studies a local stochastic volatility model in which the variance factor follows a generalized Langevin equation with colored noise, and past returns affect variance through a memory kernel with state-dependent amplitudes. It proves existence, uniqueness, and non-explosion for kernels that are finite sums of exponentials. Under structural and nondegeneracy conditions, it shows that an exact reduced Markovian state has a smooth density, so the Dupire conditional expectation can be computed without a stochastic PDE. The paper also covers premiums under the minimal entropy martingale measure, the squared VIX, and a regularized calibration problem with bid-ask constraints.
