The day in conversation for 2026-10-20
discussion
Transcript
The discussion: Iris: The latent diffusion approach for generating joint trajectories sounds promising, but how robust is the conditioning mechanism when moving beyond the initial 30 steps? Warren: I’m less concerned with the immediate step count and more with the arbitrage awareness in that first paper. If the autoencoder is learning a representation constrained by arbitrage opportunities, does that representation hold up when volatility regimes shift rapidly? Ruth: That’s where I get stuck; we need to know if the learned surface representation is truly disentangled enough to capture regime changes without introducing spurious correlations between the surface and the return path. What metrics are they using to quantify that disentanglement? Iris: They likely rely on cross-correlation analysis between the latent vectors derived from volatility and returns, but I wonder if that captures non-linear dependencies that might manifest only in extreme market conditions. How does this compare to the dynamic smoothing technique Iris mentioned
More episodes
- The day in conversation for 2026-11-19
- Daily Summary for 2026-11-19
- The day in conversation for 2026-11-18
- Daily Summary for 2026-11-18
- 2601.05274-On the use of case estimate and transactional payment data in neural networks for individual loss reserving
- 2009.00868-Loss-Given-Default Modeling by Post-Last Passage Time Process
- 2606.30070-Financial Resilience Evaluation: From Conditional Expectations to Dynamic Convex Risk Measures
- 2212.12687-Measuring price impact and information content of trades in a time-varying setting
- Daily Summary for 2026-11-17
- 2401.08323-Dynamic portfolio selection under generalized disappointment aversion