Paper: arXiv 2509.25484
Authors: Munawar Ali, Purba Das, Qi Feng, Liyao Gao, Guang Lin
Abstract
In this paper, we propose a data-driven framework for model discovery of stochastic differential equations (SDEs) from a single trajectory, without requiring the ergodicity or stationary assumption on the underlying continuous process. By combining (stochastic) Taylor expansions with Girsanov transformations, and using the drift function’s initial value as input, we construct drift estimators while simultaneously recovering the model noise. This allows us to recover the underlying \(\mathbb P\) Brownian motion increments. Building on these estimators, we introduce the first stochastic Sparse Identification of Stochastic Differential Equation (SSISDE) algorithm, capable of identifying the governing SDE dynamics from a single observed trajectory without requiring ergodicity or stationarity. To validate the proposed approach, we conduct numerical experiments with both linear and quadratic drift-diffusion functions. Among these, the Black-Scholes SDE is included as a representative case of a system that does not satisfy ergodicity or stationarity.
Complexity vs Empirical Score
- Math Complexity: 9.0/10
- Empirical Rigor: 6.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper is mathematically dense, utilizing advanced stochastic calculus (Itô/Stratonovich calculus, Girsanov transformation, stochastic Taylor expansions, rough path theory) and requires a strong theoretical foundation. While it includes numerical experiments (e.g., Black-Scholes SDE) to validate the algorithm, it focuses on model discovery and estimation from a single trajectory rather than portfolio backtesting, transaction costs, or live trading data.
Research Flowchart
flowchart TD
A["Research Goal<br>Estimate SDE noise & drift<br>from a single trajectory<br>without ergodicity"] --> B{"Data Input"}
B --> C["Single Observed Trajectory<br>e.g., Black-Scholes Market Data"]
B --> D["Drift Function Initial Value"]
C & D --> E["Methodology<br>Stochastic Taylor Expansion<br>+ Girsanov Transformation"]
E --> F["Computational Process<br>Stochastic Sparse Identification<br>of SDE SSISDE"]
F --> G["Key Outcomes"]
subgraph G
G1["Recovered Drift Function"]
G2["Estimated Model Noise<br>Recovering P Brownian Motion"]
G3["Validated on Non-Ergodic Systems<br>e.g., Black-Scholes"]
end
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a data‑driven framework for discovering stochastic differential equations from a single trajectory without ergodicity or stationarity assumptions, and introduces the Stochastic Sparse Identification of Stochastic Differential Equation (SSISDE) algorithm to identify the governing S
Source sentence
In this paper, we propose a data-driven framework for model discovery of stochastic differential equations (SDEs) from a single trajectory, without requiring the ergodicity or stationary assumption on the underlying continuous process. Building on these estimators, we introduce the first stochastic Sparse Identification of Stochastic Differential Equation (SSISDE) algorithm, capable of identifying the governing SDE
Assumes 4
- The drift µ and diffusion σ are Lipschitz continuous. model
Source sentence
1. µ and σ are Lipschitz continuous.
- µ belongs to C³(ℝ) and σ belongs to C⁴(ℝ). model
Source sentence
2. µ ∈C3(R) and σ ∈C4(R). 3.
- The initial value of the drift µ(X₀) is known (no prior knowledge of σ(X₀) required). data
Source sentence
Assume the initial values of µ i.e. µ(X0) is known (we do not need prior knowledge of σ(X0)). 4.
- The physical Brownian motion starts at zero, i.e., Bᴾ₀ = 0. model
Source sentence
Assume the P Brownian motion starts at 0 i.e. BP 0 = 0.
Data
- Synthetic linear and quadratic drift‑diffusion experiments synthetic unknown
Source sentence
To validate the proposed approach, we conduct numerical experiments with both linear and quadratic drift-diffusion functions. Among these, the Black-Scholes SDE is included as a representative case of a system that does not satisfy ergodicity or stationarity.
Costs
Not applicable
Finds 3
- The method recovers the drift function with very small mean‑squared error (MSE = 5.244 × 10⁻⁹) in the linear case. statistical
Source sentence
The overall fit is strong, with a small mean squared error MSE = 5.244 × 10−9.
- In the quadratic case the recovered drift and diffusion achieve MSE = 1.448 × 10⁻⁷, closely matching the true polynomial forms. statistical
Source sentence
The overall fit remains strong with a small mean squared error MSE = 1.448 × 10−7.
- The recovered noise closely follows the true driving signal, and simulated paths using the estimated drift and diffusion are almost indistinguishable from the observed trajectory. performance
Source sentence
The recovered noise follows the driving signal closely, and the path generated using bµ and bσ is almost indistinguishable from the observed trajectory; see Fig. 11(c)–(d). Moreover, the recovered noise follows the true driving noise well, and the path simulated using bµ and bσ is almost indistinguishable from the observed trajectory; see Fig. 18(c)–(d).
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 11 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.