Same Error, Different Function: The Optimizer as an Implicit Prior in Financial Time Series
cs.LG, q-fin.CP
Submitted: 2026-03-03
Updated: 2026-10-27
Comments: 56 pages, 38 figures
Project page: https://berts-workshop.github.io
License: http://creativecommons.org/licenses/by/4.0/
The gist: Neural networks applied to financial time series operate in a regime of underspecification, where predictors achieve comparable out-of-sample error.
Terminology
Abstract
Neural networks applied to financial time series operate in a regime of underspecification, where predictors achieve comparable out-of-sample error. Using large-scale volatility forecasting for S&P 500 stocks--twelve architecture-optimizer configurations, thirteen seeds each, evaluated on 564,795 held-out observations--we show that comparable test loss conceals measurably different predictors. In eleven of twelve comparisons, models trained from identical initialization and identical data order diverge 1.3 to 7.8 times further than repeated runs of either optimizer alone, ruling out initialization noise as the explanation. On a common scale, predictions differ by 5--17% of the typical forecast error while the models' error magnitudes differ by less than 0.7% of it. The divergence reshapes nonlinear response profiles, temporal dependence, and ultimately decisions: volatility-ranked portfolios trace a Sharpe-turnover frontier with up to 58% turnover dispersion at comparable gross Sharpe ratios. In underspecified settings, optimization is a consequential source of inductive bias, and model evaluation must extend beyond scalar loss to functional and decision-level behavior.
Sources
- Optuna: A Next-generation Hyperparameter Optimization Framework
- Edge of Stochastic Stability: Revisiting the Edge of Stability for SGD
- Invariant Risk Minimization
- On the Trajectories of SGD Without Replacement
- How Neural Networks Learn the Support is an Implicit Regularization Effect of SGD
- On the Opportunities and Risks of Foundation Models
- How Memory in Optimization Algorithms Implicitly Modifies the Loss
- The Effect of Mini-Batch Noise on the Implicit Bias of Adam
- On the Implicit Bias of Adam
- On Lazy Training in Differentiable Programming
- Adaptive Gradient Methods at the Edge of Stability
- Understanding Optimization in Deep Learning with Central Flows
- Self-Stabilization: The Implicit Bias of Gradient Descent at the Edge of Stability
- Underspecification Presents Challenges for Credibility in Modern Machine Learning
- A decoder-only foundation model for time-series forecasting
- Convergence rates for the Adam optimizer
- Sharpness-Aware Minimization for Efficiently Improving Generalization
- Shortcut Learning in Deep Neural Networks
- MOMENT: A Family of Open Time-series Foundation Models
- Implicit Regularization in Matrix Factorization
Related papers
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis
- (Mis)generalization of Helpful-Only Fine-tuning