########################################################
Code for "A Representer Theorem for Hawkes Processes
via Penalized Least Squares Minimization"
Submitted to ICLR2026.
########################################################

Step 1:
Install kernelhawkes package (Bonnet & Sangnier, AISTATS-2025).

$ pip install https://github.com/msangnier/kernelhawkes/archive/master.zip

Step 2:
Run the script to generate synthetic data used in our paper.
Generated data will be included in data_synthetic/.
File name "3D_EX_T****.dill" means the mutually-exciting scenario with horizon ****.
File name "3D_T****.dill" means the refractory scenario with horizon ****.

$ python make_synthetic_data.py

Step3:
Run the script to execute estimation experiments.
Estimation results will be included in result/, where "perf.dill" includes the
integrated squared error (ise) and CPU (cpu) scores, and "***.pdf" is a figure
which displays the estimation result. 
A dataset used can be specified by rewriting "dfile" variable at line 26 in the script.

$ python estimate_synthetic.py


