Paper: arXiv 2401.09361
Authors: Timothée Fabre, Ioane Muni Toke
Abstract
We propose a novel approach to marked Hawkes kernel inference which we name the moment-based neural Hawkes estimation method. Hawkes processes are fully characterized by their first and second order statistics through a Fredholm integral equation of the second kind. Using recent advances in solving partial differential equations with physics-informed neural networks, we provide a numerical procedure to solve this integral equation in high dimension. Together with an adapted training pipeline, we give a generic set of hyperparameters that produces robust results across a wide range of kernel shapes. We conduct an extensive numerical validation on simulated data. We finally propose two applications of the method to the analysis of the microstructure of cryptocurrency markets. In a first application we extract the influence of volume on the arrival rate of BTC-USD trades and in a second application we analyze the causality relationships and their directions amongst a universe of 15 cryptocurrency pairs in a centralized exchange.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper involves advanced mathematics including Fredholm integral equations, PDEs solved with physics-informed neural networks, and high-dimensional non-parametric estimation. It demonstrates strong empirical rigor through extensive numerical validation on simulated data and two specific applications on high-frequency cryptocurrency market data, though it lacks live trading results or deployment code.
Research Flowchart
flowchart TD
A["Research Goal: Non-Parametric<br>Hawkes Kernel Inference<br>in High Dimensions"] --> B["Methodology: Moment-Based<br>Neural Hawkes Estimation"]
B --> C["Key Component:<br>Physics-Informed Neural Networks<br>solve Fredholm Integral Equations"]
B --> D["Key Component:<br>Adapted Training Pipeline<br>for Robust Kernel Shapes"]
C & D --> E["Inputs: Simulated Data<br>for Validation & Cryptocurrency Market Data<br>BTC-USD + 15 Pairs"]
E --> F{"Computational Process"}
F --> G["High-Dimensional<br>Kernel Inference"]
G --> H["Key Findings & Outcomes"]
H --> I["Validation: Robust results<br>across diverse kernel shapes"]
H --> J["Application 1: Extracted<br>Volume influence on BTC-USD<br>arrival rates"]
H --> K["Application 2: Analyzed causality<br>directions amongst 15 crypto pairs"]
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 novel moment-based neural Hawkes estimation method for marked Hawkes kernel inference.
Source sentence
We propose a novel approach to marked Hawkes kernel inference which we name the moment-based neural Hawkes estimation method.
Assumes 3
- Non-zero marks are independent and identically distributed. model
Source sentence
We assume the non-zero marks are i.i.d.
- The branching ratio R must be less than 1 to ensure stationarity of the Hawkes process. statistical
Source sentence
The condition R < 1 ensures the stationarity of N
- The linear marked Hawkes process model assumes stationarity of the data generating process. market
Source sentence
The linear marked Hawkes process model and the second order characterization of Theorem 1 rely on the assumption of stationarity in the data generating process.
Data
- Cryptocurrency trade data crypto-onchain commercial
Source sentence
In this section we apply our estimation method to cryptocurrency trade data provided by SUN ZU Lab. The dataset is composed of trades executed between December 1st, 2023 and December 30th, 2023 on Coinbase for 15 cryptocurrency pairs.
Costs
Not applicable
Finds 3
- The estimated branching ratio on the restricted intraday window is 78%, close to the 80% obtained on the full sample, indicating robustness to seasonality. statistical
Source sentence
We find that the estimated branching ratio of the system is 78%, which is close to the one obtained when the estimation is performed over the entire period of time (80%).
- The neural Hawkes method provides much more stable kernel estimates than the Wiener‑Hopf benchmark. performance
Source sentence
We now design a simple experiment to show our non-parametric estimation method can lead to much more stable solutions than the Wiener-Hopf method.
- The method is robust to hyperparameter choices, with small batch sizes and moderate network sizes performing well. statistical
Source sentence
All in all, it is very satisfying to observe that the moment-based neural Hawkes estimation method appears robust with respect to the choice of hyperparameters.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 35 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.