The day in conversation for 2026-10-16
discussion
Transcript
The discussion: Ruth: That 8.51% error rate for N-BEATS after 365 days is compelling, especially when compared directly against the statistical baseline you mentioned; what do you think limits that generalization if we extrapolate that model into, say, non-linear price shifts rather than just temporal trends? Warren: I worry about applying it outside the established domain of second-hand electronics; while N-BEATS shows consistency across those long horizons, how robust is that consistency if the underlying demand structure changes—for instance, if a major new competitor enters the market suddenly? That relates to how we handled volatility in the Q3 forecasting models last week. Iris: The strength lies in its ability to capture multi-horizon dependence simultaneously, which is something traditional ARIMA setups struggle with; but I'm curious about the input features. Did they standardize or normalize those feature sets across different electronic categories,
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